Upload
vulien
View
214
Download
0
Embed Size (px)
Citation preview
THE JOURNAL OF FINANCE bull VOL LXXI NO 2 bull APRIL 2016
Indirect Incentives of Hedge Fund Managers
JONGHA LIM BERK A SENSOY and MICHAEL S WEISBACHlowast
ABSTRACT
Indirect incentives exist in the money management industry when good current per-formance increases future inflows of capital leading to higher future fees For theaverage hedge fund indirect incentives are at least 14 times as large as direct in-centives from incentive fees and managersrsquo personal stakes in the fund Combiningdirect and indirect incentives manager wealth increases by at least $039 for a $1increase in investor wealth Younger and more scalable hedge funds have strongerflow-performance relations leading to stronger indirect incentives These results havea number of implications for our understanding of incentives in the asset managementindustry
HEDGE FUND MANAGERS ARE AMONG the most highly paid individuals todayKaplan and Rauh (2010) estimate that in 2007 the top five hedge fund man-agers earned more than all SampP 500 firmsrsquo CEOs combined The payoff tobecoming a top hedge fund manager is therefore enormous The logic of Holm-strom (1982) Berk and Green (2004) and Chung et al (2012) provides a frame-work for understanding hedge fund managersrsquo careers Investors allocate capi-tal to funds based on their perception of managersrsquo abilities which is a functionof fund performance Good performance increases a managerrsquos lifetime incomedirectly through contractual incentive fees earned at the time of performanceIt also increases a managerrsquos lifetime income indirectly through higher futurefees both from increased flows of new investment to the fund and from themechanical increase in the fundrsquos asset base The extremely high level of payfor top hedge fund managers thus suggests that indirect incentives are likelyto be a significant component of managersrsquo total incentives particularly earlyin a managerrsquos career
lowastJongha Lim is with California State University at Fullerton Berk A Sensoy is with Ohio StateUniversity Michael S Weisbach is with Ohio State University NBER and SIFR We thank NengWang for graciously providing code for evaluating the Lan Wang and Yang (2013) model AndreaRossi provided unusually diligent research assistance For helpful comments and discussions wethank Jack Bao Jonathan Berk Niki Boyson Yawen Jiao Michael OrsquoDoherty Tarun RamadoraiJosh Rauh Ken Singleton Luke Taylor Sterling Yan two anonymous referees an anonymousAssociate Editor and seminar and conference participants at the 2014 AFA meetings the 2014Spring NBER Corporate Finance Meeting American University Arizona State University Cali-fornia State University at Fullerton Fordham University Harvard Business School NortheasternUniversity Ohio State University the University of Florida the University of Missouri and theUniversity of Oklahoma We have read The Journal of Financersquos disclosure policy and have noconflicts of interest to disclose
DOI 101111jofi12384
871
872 The Journal of Finance Rcopy
In this paper we estimate the magnitude of these indirect incentives of hedgefund managers In particular we address the following questions For an incre-mental percentage point of returns to investors how much additional capitaldoes the market allocate to that hedge fund How much of this additional cap-ital do hedge fund managers end up receiving as compensation in expectationHow does this ldquoexpected future pay for todayrsquos performancerdquo compare in mag-nitude with the direct incentive fees that hedge fund managers earn from theincremental returns How do these effects differ across types of funds and overtime for a particular fund And what are the implications of the existence ofsuch indirect incentives for hedge fund investors and for our understanding ofcontracting more broadly
We first estimate the relation between hedge fund performance and inflowsto the fund using a sample of 2998 hedge funds from 1995 to 2010 As predictedby learning models of fund allocation and consistent with prior work on mutualfunds and private equity funds this relation is substantially stronger for newerfunds whose managersrsquo abilities the market knows with less certainty For theaverage fund the estimates imply that a one-percentage-point incrementalreturn in a given quarter leads to a 15 increase in the fundrsquos assets undermanagement (AUM) from inflows of new investment over the next three yearsFor a new fund the same incremental return results in a 21 increase inAUM from inflows In addition performance has a greater impact on flows forfunds engaged in more scalable strategies These results are consistent withthe view that investors continually update their assessment of managers andadjust their portfolios accordingly
The way in which the inflow-performance relation affects managersrsquo com-pensation depends on the fee structure in hedge funds Typically hedge fundmanagers receive a management fee equal to 15 of AUM together with in-centive fees equal to 20 of profits above a high-water mark (HWM) Goodperformance increases managersrsquo future income because fees will be earnedon inflows of new investment and also because the asset value of existinginvestors becomes larger and closer to the HWM Valuing a managerrsquos com-pensation requires a contingent claims modeling framework to account for thefact that incentive fees are effectively a portfolio of call options on the fundrsquosassets We use four such models which allows us to evaluate the sensitivity ofthe estimates to different modeling frameworks and parameter choices
The first model that we use is the model of Goetzmann Ingersoll and Ross(2003 hereafter GIR) GIR provide an analytical formula for calculating thefraction of a dollar invested in the fund that in expectation will be receivedby the fundrsquos managers over the life of the fund The other three models in-corporate two real-world features that are missing from the GIR model andcould have a material impact on a managerrsquos future compensation futureperformance-based flows and the managerrsquos endogenous use of leverage inthe fundrsquos portfolio Each of these features leads to greater compensation andhence greater indirect incentives than would otherwise be the case The GIRestimates therefore provide a lower bound on the magnitude of the indirectincentives faced by hedge fund managers
Indirect Incentives of Hedge Fund Managers 873
The second model that we use augments the GIR model to accommodatefuture performance-based flows The third model is that of Lan Wang andYang (2013 hereafter LWY) in which the manager can endogenously choose theamount of leverage to use at each point in time Finally we present estimatesusing an extension of the LWY model that allows for performance-based flowsas well as endogenous leverage LWY nests GIR which assumes no leverageat any time as a special case so all of our estimates can be thought of asimplications of different versions of the LWY model
Each of these models provides an estimate of the present value of man-agersrsquo compensation per dollar invested in the fund Together with the flow-performance relations these estimates allow us to calculate the magnitude ofindirect incentives facing hedge fund managers For an incremental percent-age point or dollar of current returns to the fundrsquos investors we calculate thepresent value of the additional lifetime income the fundrsquos managers receive inexpectation due to inflows of new investment and the increase in the value ofexisting investorsrsquo assets
As a benchmark for assessing the importance of this indirect pay for perfor-mance we compare its magnitude to the direct performance pay that managersreceive from incentive fees and changes in the value of their own investment inthe fund We use the Agarwal Daniel and Naik (2009) framework to estimatethe change in the value of managersrsquo current compensation (coming from bothincentive fees and the managerrsquos own stake in the fund) for an incrementalreturn
Our estimates indicate that a one-percentage-point increase in returns gen-erates on average $331000 in expected direct incentive pay consisting of$142000 in incremental incentive fees and $190000 in incremental profits onmanagersrsquo personal stakes Using the GIR model with parameter choices thatyield lower-bound estimates we calculate that managers also receive $531000in expected future fee income consisting of $248000 in future fees earned onthe new investment that occurs in response to the incremental performanceand $282000 in extra future fees earned on the increase in the value of exist-ing investorsrsquo assets in the fund Because a one-percentage-point increase inreturn is equal to $211 million for an average-sized fund in our data thesecalculations imply that on average managers receive 16 cents in direct pay andat least 23 cents in indirect pay for each incremental dollar earned for fundinvestors The indirect career-based incentive effect is thus at least 14 timeslarger than the direct income that managers receive from incentive fees andreturns on their personal investments Again this indirect-to-direct incentiveratio of 14 corresponds to model and parameter choices that lead to a lowerbound of estimates of indirect incentives Under other plausible choices theratio would be substantially higher The average indirect-to-direct incentiveratio is 35 across all the models and parameter values we consider
We next find that indirect incentives are even larger for young funds Fornew funds we estimate indirect incentives to be six to 12 times as large asdirect incentives given the parameters used in LWY The importance of indirectincentives declines monotonically as a fund ages largely as a consequence of
874 The Journal of Finance Rcopy
weakening flow-performance relations but continues to be larger than directincentives until the fund is at least 15 years old The importance of indirectincentives also depends on the style of the fund For an average fund followinga style unlikely to be capacity-constrained indirect incentives are 32 to 73times as large as direct incentives while they are 25 to 60 times as large fora fund that is likely to be constrained and hence unable to grow as much inresponse to good performance
Overall our estimates suggest that regardless of the choice of model orreasonable model parameters the total incentives facing hedge fund managersare substantial and much larger than direct incentives alone would implyAlthough direct incentives are themselves substantial indirect incentives inthe hedge fund industry comprise the majority of managersrsquo total incentives
These estimates of substantial indirect incentives in the hedge fund indus-try have a number of implications First they are potentially important forunderstanding hedge fund contracting Hedge fund management contracts arestructured in a sophisticated manner yet perhaps surprisingly we find no ev-idence that direct compensation schemes adjust with the indirect incentivesthat their managers face The lack of such adjustment reflects a larger puzzlein our understanding of markets for alternative assets in that important con-tractual parameters most notably the 20 incentive fee vary little both acrossasset classes and across funds within asset classes
Second institutional investors often state that the financial incentives ofa potential asset manager are an important consideration when deciding be-tween alternative managers Presumably all of a managerrsquos incentives in-cluding both direct and indirect incentives matter in making this choice Theresults discussed above provide estimates of how indirect and total incentivesvary across types of funds and across similar funds of different ages whichshould be relevant for potential investors In addition indirect incentives varysystematically across types of potential investments The results here and inChung et al (2012) provide estimates of indirect and total incentives for hedgefunds and private equity funds Below we also provide estimates of indirectincentives for a sample of 11911 actively managed equity mutual funds overthe period 1995 to 2010 These estimates suggest that indirect incentives inmutual funds are substantial but smaller than those for hedge funds rangingbetween 28 and 66 of the hedge fund estimates
Finally since Fama (1980) and Holmstrom (1982) incentives generated frommanagerial career concerns have been an important part of the theory of thefirm However there are virtually no estimates of their magnitude The esti-mates provided here for hedge fund managers are among the first attempts tomeasure the importance of indirect incentives The fact that career-generatedincentives are so powerful in this industry suggests that they could be equallyimportant in other industries in which they are likely to be harder to estimate
This paper proceeds as follows Section I discusses how we quantify thedirect and indirect components of hedge fund pay for performance SectionII describes the data Section III presents estimates of the flow-performance
Indirect Incentives of Hedge Fund Managers 875
relation Section IV estimates managersrsquo direct and indirect incentives SectionV discusses implications of our estimates Section VI concludes
I Quantifying the Magnitude of Pay for Performance of HedgeFund Managers
A Direct Pay for Performance
Hedge fund managersrsquo compensation generally consists of management feeswhich are a percentage of AUM (often around 15) plus incentive fees whichare a percentage (usually 20) of profits or of profits earned above the HWMIn addition hedge fund managers usually make a personal investment in thefund The direct pay for performance that a manager receives comes from theincentive fees as well as his personal investment in the fund both of whichincrease in value with the fundrsquos performance Quantifying these direct perfor-mance incentives is complicated because of the option-like features containedin the hedge fund managerrsquos incentive fee contract In particular the incen-tive fee contract resembles a portfolio of call options one per investor in thefund The exercise price of each option is determined by the investorrsquos time ofentrance into the fund the fundrsquos hurdle rate and the historical HWM levelpertaining to the investorrsquos assets Thus even if different managers have thesame 20 incentive fee rate their actual direct pay-performance sensitivitywill vary depending on the distance between the current asset value and theexercise prices of these options
To estimate the direct pay-performance sensitivity we use Agarwal Danieland Noikrsquos (2009) total delta approach which measures the impact of an in-cremental one-percentage-point return to fund investors on the value of themanagerrsquos incentive fees plus the increase in the value of the managerrsquos ownownership stake The total delta of the managerrsquos claim to the fund is equalto the sum of these individual optionsrsquo deltas plus the delta of the managerrsquospersonal stake in the fund We follow Agarwal Daniel and Naikrsquos (2009) andassume that the managerrsquos initial stake is zero but the stake grows over timeas managers reinvest all of their incentive fees back into the fund1 Details ofthis calculation are described in Appendix A
B Indirect Pay for Performance
In addition to the pay for performance from incentive fees and their owninvestment in the fund hedge fund managersrsquo lifetime income changes withperformance through a reputational effect good performance increases themarketrsquos perception of a managerrsquos ability leading to higher inflows of newinvestment to the fund Ultimately a fundrsquos managers will receive part of
1 Assuming instead that the managerrsquos initial stake is 1 or 2 of AUM increases the estimatesof indirect incentives by 5 and 10 respectively and the estimates of direct incentives by 2and 5 respectively These changes do not affect our conclusion that indirect incentives are largerelative to direct incentives
876 The Journal of Finance Rcopy
these inflows as future management and incentive fees Furthermore goodperformance mechanically increases the value of existing investorsrsquo stakes inthe fund A portion of this increase will likewise be paid over time in future feesto the fund manager The expectation of this future income depends on todayrsquosperformance leading to what we refer to as indirect incentives
To evaluate the magnitude of these indirect incentives we first need esti-mates of the way in which performance affects expected inflows to the fundThese estimates are discussed in Section III below We also need a model ofthe present value of a managerrsquos expected lifetime compensation as a fractionof fund assets This model should predict for each incremental dollar undermanagement the increase in the managerrsquos expected compensation over thefuture lifetime of the fund We use four such models
The first model is that of GIR The GIR model leads to the lowest estimatesof the sensitivity of future income to todayrsquos performance so it can be thoughtof as providing a lower bound on our estimates of indirect incentives The GIRmodel is a contingent-claims model of the fraction of fund assets that accrue tothe fund manager in expected future compensation Key features of this modelare that compensation is a fixed management fee plus a percentage of profitsabove a HWM the fundrsquos asset value follows a martingale with drift generatedby the managerrsquos alpha investors continuously withdraw assets (eg an en-dowment investor might withdraw 5 per year) and an investor liquidates hisor her position following a sufficiently negative return shock (so that expectedfund life is finite)
The GIR model does not account for all factors that could affect managersrsquofuture compensation and in turn indirect incentives In particular the GIRmodel does not allow the fundrsquos asset value to grow through future performance-based fund flows Under appropriate assumptions the effect of such flows inthe context of the GIR model is to increase the variance of the fundrsquos AUM2
For this reason our second model reports estimates from a variant of the GIRmodel in which fund variance is augmented
Another factor not accounted for by the GIR model is the fact that a fundrsquosportfolio is endogenously chosen with managers adjusting their portfolios tomaximize their incomes given a particular incentive scheme Following LWYour third model allows managers to perform such adjustments by levering theirfunds strategically Given that hedge fund managers typically do use leverageestimates of indirect incentives that incorporate this feature are likely to bemore accurate
Finally we present estimates from a fourth model also introduced by LWYthat augments the basic LWY model to include future performance-based fundflows Note also that the LWY model nests the GIR model as a special case in
2 Suppose as an approximation that the flow-performance relationship is linear in logs andflows are contemporaneous with returns Suppose that without flows the log AUM of the fundevolves as a martingale as in the GIR model st+1 = st + et+1 With performance-based flows st+1 =st + get+1 where g gt 1 captures the flow effect Therefore even with performance-based flows thelog AUM would still follow a martingale so the GIR model still applies but with a higher variancethan in the case of no flows We thank the Associate Editor for suggesting this argument to us
Indirect Incentives of Hedge Fund Managers 877
which leverage is zero so all four sets of estimates can be viewed as comingfrom variations of the LWY model
Using different models to estimate indirect incentives allows us to gaugethe importance of different factors especially the importance of futureperformance-based fund flows and endogenous portfolio choice in determin-ing the fraction of a fundrsquos value that will ultimately accrue to managers infuture fees3 Details on all four models our choices of model parameters andthe associated calculations are described in Appendix B
For each of the models above we estimate the present value of the total(management plus incentive) future fees that the manager earns on an extradollar of AUM To calculate indirect pay for performance we multiply thispresent value by an estimate of the number of incremental dollars of AUMthat result from a one-percentage-point increase in returns to investors Thelatter increase consists of two parts namely the mechanical increase in thevalue of existing investorsrsquo stakes plus incremental inflows of new investmentIn this way the models for a managerrsquos fee value together with our estimatesof the flow-performance relation facing hedge fund managers provide an esti-mate of the present value of the incremental future revenue that the hedgefund manager expects to earn as a result of a one-percentage-point improve-ment in current net returns4 This calculation results in an incentive measuredenominated in dollars per percentage point change in returns We also con-duct the same calculation for a one-dollar improvement in investor return toobtain a measure in terms of dollars in managerial wealth per dollar returnedto shareholders Both of these measures are commonly used in the literatureon manager incentives
Because the GIR and LWY models estimate present values no further ad-justment for the riskiness of future income is required Also the estimates donot require that the manager continue to manage the fund in the future underthe assumption that the present value of the managerrsquos claims to future feeincome can be monetized when the manager departs
II Hedge Fund Data
Our data come from the TASS database which covers about 40 of thehedge fund universe (Agarwal Daniel and Naik (2009)) Summary statisticsfor key sample fund characteristics reported in Table I are very close to thosefor the sample considered by Agarwal Daniel and Naik (2009) who mergeand consolidate four major databases (CISDM HFR MSCI and TASS) Wetherefore conclude that our sample of hedge funds is representative of thehedge fund universe
3 In addition to GIR and LWY there have been a number of other attempts to value managersrsquoclaims to hedge funds Important contributions include Panageas and Westerfield (2009) Drechsler(2014) and Guasoni and Oblόj (2013)
4 Net returns can be improved either through increased gross returns or decreased costs borneby the fund such as financing costs security lending fees and settlement charges The incentiveswe measure are therefore incentives to achieve both
878 The Journal of Finance Rcopy
Table ISummary Statistics
This table presents descriptive statistics for the sample of 2998 funds during the 1995 to 2010sample period Time-varying variables are reported at the fund-quarter level and other time-invariant contractual characteristics are reported at the fund level Quarterly flow is the differencebetween the reported quarter-end AUM and the quarter-beginning AUM times (1 + Quarterlyreturns) scaled by quarter-beginning AUM Quarterly returns are the reported quarterly net-of-fee returns AUM is assets under management in millions measured at the end of each quarterAge is the number of quarters from the fundrsquos inception date measured at the beginning of eachquarter Volatility is the standard deviation of monthly (net-of-fee) returns over the 12 monthsprior to each quarter multiplied by the square root of three to arrive at a quarterly figure Alltime-varying variables except Age are winsorized at the 1st and 99th percentiles Management feesare the percentage of the assets charged by the fund as regular fees in annual terms Incentive feesare the percentage of positive profits received by the manager as performance incentives HWMis an indicator variable that equals one if the fund has a HWM provision and zero otherwiseLeverage is an indicator variable that equals one if the fund uses leverage and zero otherwiseOpen-to-public is an indicator variable that equals one if the fund allows public investment andzero otherwise On-shore is an indicator variable that equals one if the fund is domiciled in theUnited States and zero otherwise Total redemption period is the sum of the notice period and theredemption period measured in quarters where the notice period is the time the investor has togive notice to the fund about an intention to withdraw money from the fund and the redemptionperiod is the time that the fund takes to return the money after the notice period is over Lockupperiod is the minimum time in quarters that an investor has to wait before withdrawing investedmoney Lockup period is considered to be zero for a fund that has no lock-up provision Subscriptionperiod is the time delay measured in quarters between investing in a fund and actually purchasingfund shares Constrained is an indicator variable that equals one if the fundrsquos strategy belongs toConvertible Arbitrage Fixed Income Arbitrage Emerging Markets and Event Driven strategiesand zero otherwise (Ding et al (2009))
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 50333 798 minus413 034 1014 3701Quarterly returns () 50333 256 minus091 211 566 905AUM ($ million) 50333 2108 110 393 1300 6370Age (quarters) 50333 1756 609 1319 2434 1532Volatility () 50333 533 300 441 804 243Management fees ( annualized) 2998 15 10 15 20 05Incentive fees () 2998 193 200 200 200 34HWM 2998 0694 0000 1000 1000 0461Leverage 2998 0685 0000 1000 1000 0464Open-to-public 2998 0191 0000 0000 0000 0394On-shore 2998 0230 0000 0000 0000 0421Total redemption period (quarters) 2998 1085 0556 0667 1333 1015Lockup period (quarters) 2998 0974 0000 0000 0000 2077Subscription period (quarters) 2998 0356 0333 0333 0333 0190Style
Constrained 2998 0303 0000 0000 1000 0460Constrained = 1
Convertible Arbitrage 2998 0034Emerging Markets 2998 0120Event Driven 2998 0096Fixed Income Arbitrage 2998 0052
Constrained = 0Equity Market Neutral 2998 0073Global Macro 2998 0078LongShort Equity Hedge 2998 0420Multi-Strategy 2998 0077Other 2998 0049
Indirect Incentives of Hedge Fund Managers 879
Our sample period is from January 1995 to December 2010 We focus onthe post-1994 period because TASS started reporting information on ldquodefunctrdquofunds only after 19945 We exclude managed futuresCTAs and funds-of-fundswhich have different incentive fees and are likely to have different inflow-performance relations from typical individual hedge funds We also excludeclosed-end hedge funds as subscriptions in these funds are only possible duringthe initial issuing period so future flows are not possible This initial filterleaves us with 4939 open-end hedge funds6
We drop funds for which TASS does not contain information on organizationalcharacteristics such as management fees incentive fees and HWM provisionsIn addition we consider only funds with an uninterrupted series of net assetvalues and returns so that we can calculate inflows Further we restrict oursample to funds with at least 12 consecutive monthly returns available duringthe sample period If a fund stops reporting returns and then resumes at afuture date we use only the first sequence of uninterrupted data Finally weexclude funds with an incentive fee of zero because there can be no direct payfor performance for these funds
We conduct the analysis using quarterly data because we wish to include onlylagged (not contemporaneous) returns in the flow-performance specificationsand to have a relatively short gap between returns and flows7 To construct aquarterly sample we drop all fund-calendar quarters that have return infor-mation only for a fraction of the period We also require that a sample fundhave a subscription period less than or equal to three months so that quarterlyinflows are not restricted This sample construction process leaves us with asample of 2998 funds (50333 fund-quarter observations)
Table I presents descriptive statistics Time-varying variables such as Quar-terly flow and Quarterly returns are measured at the fund-quarter level andother contractual characteristics such as Management fees and Incentive feesare measured at the fund level8 All time-varying variables except fund age(Age) are winsorized at the 1st and 99th percentiles to minimize the effect ofoutliers
5 Defunct funds include funds that are liquidated merged or restructured as well as those thatstopped reporting returns to TASS (Fung et al (2008))
6 Some funds are closed to new investors but unfortunately we do not know whether a particularfund is taking new money at any point in time so we cannot exclude funds on the basis of thispolicy Including closed funds causes us to understate the flow-performance relation for funds thatare not closed
7 Prior versions of this paper presented estimates using annual data with contemporaneousannual returns included in the flow-performance specifications (Lim Sensoy and Weisbach (2013))The estimates of indirect incentives using annual data are very close to those reported here
8 TASS provides information on fundsrsquo organizational characteristics as of the last available dateof fund data Like most previous studies we also assume that these organizational characteristicsdo not change throughout the life of the fund Agarwal Daniel and Naik (2009) argue that fundsrsquoorganizational characteristics are unlikely to change much over time based on their discussionswith practitioners which suggests that if a manager wants to impose new contractual terms it iseasier for him to start a new fund with different terms than to go through the legal complicationsof changing an existing contract
880 The Journal of Finance Rcopy
Mean Quarterly flow is 80 with a median of 03 so the distributionis highly skewed Mean Quarterly returns is 26 the median is 21 Aver-age fund size (AUM) is $2108 million The remaining variables reflect time-invariant contractual features Summary statistics on these characteristics arevery close to those reported in prior studies (eg Agarwal Daniel and Naik(2009) Baquero and Verbeek (2009) Aragon and Nanda (2012)) The variableManagement fees is the annual percentage of fund assets (AUM) received bythe manager as compensation and has a sample mean (median) of 15 (15)The variable Incentive fees is the percentage of profits above the HWM receivedby the manager as compensation and has a sample mean (median) of 193(20) Over two-thirds of our sample funds 694 have a HWM provision685 report that they use leverage 191 are open to public investors andabout a quarter are on-shore funds
The fee data consist of the fees that are currently publicly quoted by thefunds These data could potentially misrepresent the true fees relevant forour calculations for three reasons First funds sometimes provide fee reduc-tions to particular strategic investors that are not reflected in the database(Ramadorai and Streatfield (2011)) Although we cannot investigate this pos-sibility directly it is unlikely to have a major impact on our conclusions Forexample if the true incentive fee (management fee) averaged over all investorswere 10 lower than what is reported in the database our estimates of indirectincentives using the base GIR model would be overstated by about 1 (5) Asecond issue is that fees can change over time Agarwal and Ray (2012) andDueskar et al (2012) find that fee changes are infrequent and tend to reflectpast performance when they do occur so that fee increases follow good perfor-mance and fee decreases follow poor performance This effect is magnified inindirect incentives because good performance today leads not only to inflowsbut also to higher proportional fees on those inflows Third it is possible thatthere are fee changes unobservable to researchers To the extent that observ-able and unobservable fee changes are positively correlated this possibilityleads us to understate indirect incentives
We consider three variables that reflect potential restrictions on the be-havior of flows First we use Total redemption period defined as the sum ofthe notice period and the redemption period where the notice period is thetime the investor has to give notice to the fund about an intention to with-draw money from the fund and the redemption period is the time that thefund takes to return the money after the notice period is over The Lockupperiod is the minimum time that an investor has to wait before withdrawinginvested money The Subscription period is the time delay between investingin a fund and actually purchasing fund shares The mean total redemption pe-riod lock-up period and subscription period are 109 097 and 036 quartersrespectively
Indirect Incentives of Hedge Fund Managers 881
III Estimating the Sensitivity of Fund Inflows to Performance
To understand the impact of performance on fund flows we employ aBayesian learning framework that presumes investors are continually eval-uating managers trying to assess their abilities (see Berk and Green (2004)Chung et al (2012)) A fundrsquos performance provides information about the man-agerrsquos ability so an observation of performance will lead investors to updatetheir assessment of his ability and allocate more capital to a fund if their es-timate of the managerrsquos ability increases The magnitude of the updates andhence the sensitivity of inflows to performance should depend on the infor-mativeness of the signal relative to the precision of the prior estimate of thefund managerrsquos ability The sensitivity of inflows to performance should alsodepend on the extent to which ability can be scaled to replicate a fundrsquos returndistribution on new capital
Measuring the indirect incentives of hedge fund managers requires an es-timate of the relation between fund performance and future inflows A longliterature beginning with Ippolito (1992) estimates this relation to be rela-tively strong in the mutual fund industry9 It is also positive in the privateequity industry (see eg Kaplan and Schoar (2005)) However the results forhedge funds are less clear Goetzmann Ingersoll and Ross (2003) report a neg-ative and concave relation whereas other studies including Agarwal Danieland Naik (2003) Fung et al (2008) Baquero and Verbeek (2009) and Dinget al (2009) find a positive one
A Empirical Specification
We estimate the following specification
Flowit = β0 +11sumj=1
β0+ j Returnitminus j + γ Xtminus1 + λY + Fixed effects + εit (1)
where Flowit represents flows for fund i in quarter t10
Following the literature on flows to mutual funds (eg Sirri and Tufano(1998) Chevalier and Ellison (1999)) or hedge funds (eg Fung et al (2008)Agarwal Daniel and Naik (2009)) we compute quarterly flows of capital intoa fund as follows
Flowit = AUMit minus AUMitminus1(1 + Returnit
)AUMitminus1
(2)
where AUMit and AUMitminus1 are the AUM of fund i at the end of quarters t andtminus1 respectively and Returnit is the net-of-fee return for fund i in quarter t
9 See Brown Halow and Starks (1996) Sirri and Tufano (1998) Chevalier and Ellison (1999)Barclay Pearson and Weisbach (1998) Del Guercio and Tkac (2002) Bollen (2007) Huang Weiand Yan (2007) and Sensoy (2009)
10 We restrict our estimates to quarterly specifications but prior versions of this paper alsoinclude annual and monthly specifications with similar results to those reported below
882 The Journal of Finance Rcopy
This definition expresses flows as a fraction of beginning-of-period (end ofprior period) AUM which is a natural benchmark from the perspective of afund manager assessing his or her incentives going forward For instance theoption deltas that comprise direct incentives are defined in terms of beginning-of-period AUM An alternative approach to computing fund flows is to scalethe denominator of equation (2) by (1 + Returnit) which expresses flows as afraction of what end-of-period AUM would have been in the absence of flowsAlthough this alternative definition is less intuitive for our purpose it leads tosimilar estimates of indirect incentives
The vector X in equation (1) consists of time-varying fund characteristicsthat include lagged flows the natural logarithm of AUM for fund i at tminus1 thenatural logarithm of one plus fund irsquos age in quarters at tminus1 and fund volatil-ity over the previous 12 months11 The vector Y includes time-invariant fundcharacteristics that include Management fees Incentive fees Total redemptionperiod Lockup period Subscription period and a set of indicator variables thatequal one if the fund has an HWM provision uses leverage is open to publicinvestors and is an on-shore fund All specifications include fixed effects forthe nine styles listed in Table I interacted with calendar quarter fixed effectsThese time-by-style effects capture all shocks observed or unobserved that arecommon to funds of a given style in a given quarter including the returns topeer funds and inflows to other funds of the same style12 Reported standarderrors are robust to heteroskedasticity and account for double clustering byfund and time period13
B Estimates of the Flow-Performance Relation
We present estimates of the flow-performance relation in Panel A ofTable II In Column (1) we include returns in the 11 quarters prior to thecurrent quarter In Column (2) we add a number of fund-level controls In eachspecification the coefficients on returns are positive and statistically signif-icant in most cases and decline sharply over time so the coefficient on themost recent quarterrsquos return is the largest If we sum the coefficients on the 11prior quarters the sum in Column (1) is 144 and in Column (2) is 150 Thesecoefficients imply that a one-standard-deviation increase in returns (9) willlead to about a 13 increase in fund size
11 For young funds that have for example only one yearrsquos worth of return history we cannotcompute lagged returns and flows In such cases we ldquodummy outrdquo missing lagged variables toretain observations To do so we set missing values of lagged flows and returns to zero and includean indicator for missing values
12 Time-by-style fixed effects perform the same adjustment as a factor model regression underthe assumption that the factor loadings are the same for all funds of a given style within a giventime period
13 We focus on linear specifications Although there is some evidence of nonlinear effects in thedata (using splines quadratics etc) they are small in magnitude and have little impact on theestimates obtained with linear specifications
Indirect Incentives of Hedge Fund Managers 883
Table IIEstimates of Flow-Performance Sensitivity
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) using quarterly data The dependent variable is Quarterly flow as defined byequation (2) See Table I for definitions of all independent variables The sample period is from 1995to 2010 We only employ the funds with a subscription period less than or equal to three months toavoid quarterly inflows being restricted by subscription periods Equation (1) is augmented in PanelB to include interaction terms of the log of the fundrsquos age plus one with prior performance and inPanel C to include interaction terms of Constrained with performance variables All specificationsin Panel A and B include style fixed effects quarter fixed effects and quarter-by-style fixed effectsOnly quarter fixed effects are included in Panel C due to multicollinearity Standard errors aredouble clustered by fund and year and indicate statistical significance at the 1 5and 10 level respectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel A Base-case
Quarterly returns(tminus1) 0501 (0045) 0496 (0043)Quarterly returns(tminus2) 0314 (0039) 0263 (0036)Quarterly returns(tminus3) 0209 (0034) 0190 (0028)Quarterly returns(tminus4) 0117 (0035) 0111 (0030)Quarterly returns(tminus5) 0089 (0030) 0098 (0027)Quarterly returns(tminus6) 0054 (0033) 0066 (0030)Quarterly returns(tminus7) 0083 (0027) 0103 (0025)Quarterly returns(tminus8) 0044 (0022) 0060 (0022)Quarterly returns(tminus9) 0012 (0029) 0034 (0029)Quarterly returns(tminus10) minus0010 (0030) 0019 (0029)Quarterly returns(tminus11) 0025 (0027) 0056 (0026)Quarterly flow(tminus1) 0158 (0008)Log(Age + 1) minus0004 (0005)Log(AUM(tminus1)+1) minus0024 (0001)Volatility minus0141 (0037)Management fees 0211 (0386)Incentive fees 0005 (0053)HWM indicator 0020 (0004)Leveraged indicator 0000 (0004)Open-to-public indicator minus0002 (0005)On-shore indicator minus0032 (0005)Log(Total redemption period +1) 0034 (0007)Log(Lock-up period +1) minus0005 (0003)Log(Subscription period +1) minus0023 (0015)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0162 0194F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1940 (0000) 2077 (0000)
(Continued)
884 The Journal of Finance Rcopy
Table IImdashContinued
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel B Age interactions
Quarterly returns(tminus1)timesLog(Age in quarters+1) minus0319 (0035) minus0285 (0034)Quarterly returns(tminus2)timesLog(Age in quarters+1) minus0171 (0027) minus0125 (0025)Quarterly returns(tminus3)timesLog(Age in quarters+1) minus0051 (0036) minus0033 (0036)Quarterly returns(tminus4)timesLog(Age in quarters+1) minus0003 (0032) 0001 (0032)Quarterly returns(tminus5)timesLog(Age in quarters+1) minus0041 (0030) minus0042 (0029)Quarterly returns(tminus6)timesLog(Age in quarters+1) 0038 (0033) 0040 (0031)Quarterly returns(tminus7)timesLog(Age in quarters+1) 0013 (0033) 0007 (0032)Quarterly returns(tminus8)timesLog(Age in quarters+1) 0033 (0033) 0034 (0033)Quarterly returns(tminus9)timesLog(Age in quarters+1) 0003 (0040) minus0004 (0040)Quarterly returns(tminus10)timesLog(Age in quarters+1) 0132 (0038) 0133 (0035)Quarterly returns(tminus11)timesLog(Age in quarters+1) 0018 (0040) 0007 (0041)Quarterly returns(tminus1) 1325 (0117) 1233 (0113)Quarterly returns(tminus2) 0773 (0095) 0602 (0087)Quarterly returns(tminus3) 0354 (0114) 0287 (0110)Quarterly returns(tminus4) 0124 (0112) 0110 (0106)Quarterly returns(tminus5) 0200 (0100) 0215 (0092)Quarterly returns(tminus6) minus0067 (0112) minus0056 (0104)Quarterly returns(tminus7) 0033 (0106) 0079 (0099)Quarterly returns(tminus8) minus0066 (0107) minus0046 (0107)Quarterly returns(tminus9) 0004 (0128) 0053 (0127)Quarterly returns(tminus10) minus0416 (0122) minus0383 (0114)Quarterly returns(tminus11) minus0038 (0126) 0037 (0130)Log(Age + 1) minus0026 (0007) minus0000 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0168 0198F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1788 (0000) 1570 (0000)All age interaction terms 1214 (0000) 1084 (0000)
Panel C Scalability interactions
Quarterly returns(tminus1)timesConstrained minus0215 (0048) minus0215 (0047)Quarterly returns(tminus2)timesConstrained minus0096 (0043) minus0078 (0039)Quarterly returns(tminus3)timesConstrained minus0102 (0038) minus0107 (0037)Quarterly returns(tminus4)timesConstrained 0013 (0034) 0013 (0036)Quarterly returns(tminus5)timesConstrained minus0009 (0032) minus0022 (0032)Quarterly returns(tminus6)timesConstrained 0026 (0034) 0013 (0034)Quarterly returns(tminus7)timesConstrained minus0044 (0036) minus0057 (0036)Quarterly returns(tminus8)timesConstrained 0032 (0037) 0033 (0036)Quarterly returns(tminus9)timesConstrained minus0085 (0030) minus0090 (0031)Quarterly returns(tminus10)timesConstrained minus0023 (0036) minus0007 (0036)Quarterly returns(tminus11)timesConstrained minus0007 (0037) minus0004 (0038)
(Continued)
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
872 The Journal of Finance Rcopy
In this paper we estimate the magnitude of these indirect incentives of hedgefund managers In particular we address the following questions For an incre-mental percentage point of returns to investors how much additional capitaldoes the market allocate to that hedge fund How much of this additional cap-ital do hedge fund managers end up receiving as compensation in expectationHow does this ldquoexpected future pay for todayrsquos performancerdquo compare in mag-nitude with the direct incentive fees that hedge fund managers earn from theincremental returns How do these effects differ across types of funds and overtime for a particular fund And what are the implications of the existence ofsuch indirect incentives for hedge fund investors and for our understanding ofcontracting more broadly
We first estimate the relation between hedge fund performance and inflowsto the fund using a sample of 2998 hedge funds from 1995 to 2010 As predictedby learning models of fund allocation and consistent with prior work on mutualfunds and private equity funds this relation is substantially stronger for newerfunds whose managersrsquo abilities the market knows with less certainty For theaverage fund the estimates imply that a one-percentage-point incrementalreturn in a given quarter leads to a 15 increase in the fundrsquos assets undermanagement (AUM) from inflows of new investment over the next three yearsFor a new fund the same incremental return results in a 21 increase inAUM from inflows In addition performance has a greater impact on flows forfunds engaged in more scalable strategies These results are consistent withthe view that investors continually update their assessment of managers andadjust their portfolios accordingly
The way in which the inflow-performance relation affects managersrsquo com-pensation depends on the fee structure in hedge funds Typically hedge fundmanagers receive a management fee equal to 15 of AUM together with in-centive fees equal to 20 of profits above a high-water mark (HWM) Goodperformance increases managersrsquo future income because fees will be earnedon inflows of new investment and also because the asset value of existinginvestors becomes larger and closer to the HWM Valuing a managerrsquos com-pensation requires a contingent claims modeling framework to account for thefact that incentive fees are effectively a portfolio of call options on the fundrsquosassets We use four such models which allows us to evaluate the sensitivity ofthe estimates to different modeling frameworks and parameter choices
The first model that we use is the model of Goetzmann Ingersoll and Ross(2003 hereafter GIR) GIR provide an analytical formula for calculating thefraction of a dollar invested in the fund that in expectation will be receivedby the fundrsquos managers over the life of the fund The other three models in-corporate two real-world features that are missing from the GIR model andcould have a material impact on a managerrsquos future compensation futureperformance-based flows and the managerrsquos endogenous use of leverage inthe fundrsquos portfolio Each of these features leads to greater compensation andhence greater indirect incentives than would otherwise be the case The GIRestimates therefore provide a lower bound on the magnitude of the indirectincentives faced by hedge fund managers
Indirect Incentives of Hedge Fund Managers 873
The second model that we use augments the GIR model to accommodatefuture performance-based flows The third model is that of Lan Wang andYang (2013 hereafter LWY) in which the manager can endogenously choose theamount of leverage to use at each point in time Finally we present estimatesusing an extension of the LWY model that allows for performance-based flowsas well as endogenous leverage LWY nests GIR which assumes no leverageat any time as a special case so all of our estimates can be thought of asimplications of different versions of the LWY model
Each of these models provides an estimate of the present value of man-agersrsquo compensation per dollar invested in the fund Together with the flow-performance relations these estimates allow us to calculate the magnitude ofindirect incentives facing hedge fund managers For an incremental percent-age point or dollar of current returns to the fundrsquos investors we calculate thepresent value of the additional lifetime income the fundrsquos managers receive inexpectation due to inflows of new investment and the increase in the value ofexisting investorsrsquo assets
As a benchmark for assessing the importance of this indirect pay for perfor-mance we compare its magnitude to the direct performance pay that managersreceive from incentive fees and changes in the value of their own investment inthe fund We use the Agarwal Daniel and Naik (2009) framework to estimatethe change in the value of managersrsquo current compensation (coming from bothincentive fees and the managerrsquos own stake in the fund) for an incrementalreturn
Our estimates indicate that a one-percentage-point increase in returns gen-erates on average $331000 in expected direct incentive pay consisting of$142000 in incremental incentive fees and $190000 in incremental profits onmanagersrsquo personal stakes Using the GIR model with parameter choices thatyield lower-bound estimates we calculate that managers also receive $531000in expected future fee income consisting of $248000 in future fees earned onthe new investment that occurs in response to the incremental performanceand $282000 in extra future fees earned on the increase in the value of exist-ing investorsrsquo assets in the fund Because a one-percentage-point increase inreturn is equal to $211 million for an average-sized fund in our data thesecalculations imply that on average managers receive 16 cents in direct pay andat least 23 cents in indirect pay for each incremental dollar earned for fundinvestors The indirect career-based incentive effect is thus at least 14 timeslarger than the direct income that managers receive from incentive fees andreturns on their personal investments Again this indirect-to-direct incentiveratio of 14 corresponds to model and parameter choices that lead to a lowerbound of estimates of indirect incentives Under other plausible choices theratio would be substantially higher The average indirect-to-direct incentiveratio is 35 across all the models and parameter values we consider
We next find that indirect incentives are even larger for young funds Fornew funds we estimate indirect incentives to be six to 12 times as large asdirect incentives given the parameters used in LWY The importance of indirectincentives declines monotonically as a fund ages largely as a consequence of
874 The Journal of Finance Rcopy
weakening flow-performance relations but continues to be larger than directincentives until the fund is at least 15 years old The importance of indirectincentives also depends on the style of the fund For an average fund followinga style unlikely to be capacity-constrained indirect incentives are 32 to 73times as large as direct incentives while they are 25 to 60 times as large fora fund that is likely to be constrained and hence unable to grow as much inresponse to good performance
Overall our estimates suggest that regardless of the choice of model orreasonable model parameters the total incentives facing hedge fund managersare substantial and much larger than direct incentives alone would implyAlthough direct incentives are themselves substantial indirect incentives inthe hedge fund industry comprise the majority of managersrsquo total incentives
These estimates of substantial indirect incentives in the hedge fund indus-try have a number of implications First they are potentially important forunderstanding hedge fund contracting Hedge fund management contracts arestructured in a sophisticated manner yet perhaps surprisingly we find no ev-idence that direct compensation schemes adjust with the indirect incentivesthat their managers face The lack of such adjustment reflects a larger puzzlein our understanding of markets for alternative assets in that important con-tractual parameters most notably the 20 incentive fee vary little both acrossasset classes and across funds within asset classes
Second institutional investors often state that the financial incentives ofa potential asset manager are an important consideration when deciding be-tween alternative managers Presumably all of a managerrsquos incentives in-cluding both direct and indirect incentives matter in making this choice Theresults discussed above provide estimates of how indirect and total incentivesvary across types of funds and across similar funds of different ages whichshould be relevant for potential investors In addition indirect incentives varysystematically across types of potential investments The results here and inChung et al (2012) provide estimates of indirect and total incentives for hedgefunds and private equity funds Below we also provide estimates of indirectincentives for a sample of 11911 actively managed equity mutual funds overthe period 1995 to 2010 These estimates suggest that indirect incentives inmutual funds are substantial but smaller than those for hedge funds rangingbetween 28 and 66 of the hedge fund estimates
Finally since Fama (1980) and Holmstrom (1982) incentives generated frommanagerial career concerns have been an important part of the theory of thefirm However there are virtually no estimates of their magnitude The esti-mates provided here for hedge fund managers are among the first attempts tomeasure the importance of indirect incentives The fact that career-generatedincentives are so powerful in this industry suggests that they could be equallyimportant in other industries in which they are likely to be harder to estimate
This paper proceeds as follows Section I discusses how we quantify thedirect and indirect components of hedge fund pay for performance SectionII describes the data Section III presents estimates of the flow-performance
Indirect Incentives of Hedge Fund Managers 875
relation Section IV estimates managersrsquo direct and indirect incentives SectionV discusses implications of our estimates Section VI concludes
I Quantifying the Magnitude of Pay for Performance of HedgeFund Managers
A Direct Pay for Performance
Hedge fund managersrsquo compensation generally consists of management feeswhich are a percentage of AUM (often around 15) plus incentive fees whichare a percentage (usually 20) of profits or of profits earned above the HWMIn addition hedge fund managers usually make a personal investment in thefund The direct pay for performance that a manager receives comes from theincentive fees as well as his personal investment in the fund both of whichincrease in value with the fundrsquos performance Quantifying these direct perfor-mance incentives is complicated because of the option-like features containedin the hedge fund managerrsquos incentive fee contract In particular the incen-tive fee contract resembles a portfolio of call options one per investor in thefund The exercise price of each option is determined by the investorrsquos time ofentrance into the fund the fundrsquos hurdle rate and the historical HWM levelpertaining to the investorrsquos assets Thus even if different managers have thesame 20 incentive fee rate their actual direct pay-performance sensitivitywill vary depending on the distance between the current asset value and theexercise prices of these options
To estimate the direct pay-performance sensitivity we use Agarwal Danieland Noikrsquos (2009) total delta approach which measures the impact of an in-cremental one-percentage-point return to fund investors on the value of themanagerrsquos incentive fees plus the increase in the value of the managerrsquos ownownership stake The total delta of the managerrsquos claim to the fund is equalto the sum of these individual optionsrsquo deltas plus the delta of the managerrsquospersonal stake in the fund We follow Agarwal Daniel and Naikrsquos (2009) andassume that the managerrsquos initial stake is zero but the stake grows over timeas managers reinvest all of their incentive fees back into the fund1 Details ofthis calculation are described in Appendix A
B Indirect Pay for Performance
In addition to the pay for performance from incentive fees and their owninvestment in the fund hedge fund managersrsquo lifetime income changes withperformance through a reputational effect good performance increases themarketrsquos perception of a managerrsquos ability leading to higher inflows of newinvestment to the fund Ultimately a fundrsquos managers will receive part of
1 Assuming instead that the managerrsquos initial stake is 1 or 2 of AUM increases the estimatesof indirect incentives by 5 and 10 respectively and the estimates of direct incentives by 2and 5 respectively These changes do not affect our conclusion that indirect incentives are largerelative to direct incentives
876 The Journal of Finance Rcopy
these inflows as future management and incentive fees Furthermore goodperformance mechanically increases the value of existing investorsrsquo stakes inthe fund A portion of this increase will likewise be paid over time in future feesto the fund manager The expectation of this future income depends on todayrsquosperformance leading to what we refer to as indirect incentives
To evaluate the magnitude of these indirect incentives we first need esti-mates of the way in which performance affects expected inflows to the fundThese estimates are discussed in Section III below We also need a model ofthe present value of a managerrsquos expected lifetime compensation as a fractionof fund assets This model should predict for each incremental dollar undermanagement the increase in the managerrsquos expected compensation over thefuture lifetime of the fund We use four such models
The first model is that of GIR The GIR model leads to the lowest estimatesof the sensitivity of future income to todayrsquos performance so it can be thoughtof as providing a lower bound on our estimates of indirect incentives The GIRmodel is a contingent-claims model of the fraction of fund assets that accrue tothe fund manager in expected future compensation Key features of this modelare that compensation is a fixed management fee plus a percentage of profitsabove a HWM the fundrsquos asset value follows a martingale with drift generatedby the managerrsquos alpha investors continuously withdraw assets (eg an en-dowment investor might withdraw 5 per year) and an investor liquidates hisor her position following a sufficiently negative return shock (so that expectedfund life is finite)
The GIR model does not account for all factors that could affect managersrsquofuture compensation and in turn indirect incentives In particular the GIRmodel does not allow the fundrsquos asset value to grow through future performance-based fund flows Under appropriate assumptions the effect of such flows inthe context of the GIR model is to increase the variance of the fundrsquos AUM2
For this reason our second model reports estimates from a variant of the GIRmodel in which fund variance is augmented
Another factor not accounted for by the GIR model is the fact that a fundrsquosportfolio is endogenously chosen with managers adjusting their portfolios tomaximize their incomes given a particular incentive scheme Following LWYour third model allows managers to perform such adjustments by levering theirfunds strategically Given that hedge fund managers typically do use leverageestimates of indirect incentives that incorporate this feature are likely to bemore accurate
Finally we present estimates from a fourth model also introduced by LWYthat augments the basic LWY model to include future performance-based fundflows Note also that the LWY model nests the GIR model as a special case in
2 Suppose as an approximation that the flow-performance relationship is linear in logs andflows are contemporaneous with returns Suppose that without flows the log AUM of the fundevolves as a martingale as in the GIR model st+1 = st + et+1 With performance-based flows st+1 =st + get+1 where g gt 1 captures the flow effect Therefore even with performance-based flows thelog AUM would still follow a martingale so the GIR model still applies but with a higher variancethan in the case of no flows We thank the Associate Editor for suggesting this argument to us
Indirect Incentives of Hedge Fund Managers 877
which leverage is zero so all four sets of estimates can be viewed as comingfrom variations of the LWY model
Using different models to estimate indirect incentives allows us to gaugethe importance of different factors especially the importance of futureperformance-based fund flows and endogenous portfolio choice in determin-ing the fraction of a fundrsquos value that will ultimately accrue to managers infuture fees3 Details on all four models our choices of model parameters andthe associated calculations are described in Appendix B
For each of the models above we estimate the present value of the total(management plus incentive) future fees that the manager earns on an extradollar of AUM To calculate indirect pay for performance we multiply thispresent value by an estimate of the number of incremental dollars of AUMthat result from a one-percentage-point increase in returns to investors Thelatter increase consists of two parts namely the mechanical increase in thevalue of existing investorsrsquo stakes plus incremental inflows of new investmentIn this way the models for a managerrsquos fee value together with our estimatesof the flow-performance relation facing hedge fund managers provide an esti-mate of the present value of the incremental future revenue that the hedgefund manager expects to earn as a result of a one-percentage-point improve-ment in current net returns4 This calculation results in an incentive measuredenominated in dollars per percentage point change in returns We also con-duct the same calculation for a one-dollar improvement in investor return toobtain a measure in terms of dollars in managerial wealth per dollar returnedto shareholders Both of these measures are commonly used in the literatureon manager incentives
Because the GIR and LWY models estimate present values no further ad-justment for the riskiness of future income is required Also the estimates donot require that the manager continue to manage the fund in the future underthe assumption that the present value of the managerrsquos claims to future feeincome can be monetized when the manager departs
II Hedge Fund Data
Our data come from the TASS database which covers about 40 of thehedge fund universe (Agarwal Daniel and Naik (2009)) Summary statisticsfor key sample fund characteristics reported in Table I are very close to thosefor the sample considered by Agarwal Daniel and Naik (2009) who mergeand consolidate four major databases (CISDM HFR MSCI and TASS) Wetherefore conclude that our sample of hedge funds is representative of thehedge fund universe
3 In addition to GIR and LWY there have been a number of other attempts to value managersrsquoclaims to hedge funds Important contributions include Panageas and Westerfield (2009) Drechsler(2014) and Guasoni and Oblόj (2013)
4 Net returns can be improved either through increased gross returns or decreased costs borneby the fund such as financing costs security lending fees and settlement charges The incentiveswe measure are therefore incentives to achieve both
878 The Journal of Finance Rcopy
Table ISummary Statistics
This table presents descriptive statistics for the sample of 2998 funds during the 1995 to 2010sample period Time-varying variables are reported at the fund-quarter level and other time-invariant contractual characteristics are reported at the fund level Quarterly flow is the differencebetween the reported quarter-end AUM and the quarter-beginning AUM times (1 + Quarterlyreturns) scaled by quarter-beginning AUM Quarterly returns are the reported quarterly net-of-fee returns AUM is assets under management in millions measured at the end of each quarterAge is the number of quarters from the fundrsquos inception date measured at the beginning of eachquarter Volatility is the standard deviation of monthly (net-of-fee) returns over the 12 monthsprior to each quarter multiplied by the square root of three to arrive at a quarterly figure Alltime-varying variables except Age are winsorized at the 1st and 99th percentiles Management feesare the percentage of the assets charged by the fund as regular fees in annual terms Incentive feesare the percentage of positive profits received by the manager as performance incentives HWMis an indicator variable that equals one if the fund has a HWM provision and zero otherwiseLeverage is an indicator variable that equals one if the fund uses leverage and zero otherwiseOpen-to-public is an indicator variable that equals one if the fund allows public investment andzero otherwise On-shore is an indicator variable that equals one if the fund is domiciled in theUnited States and zero otherwise Total redemption period is the sum of the notice period and theredemption period measured in quarters where the notice period is the time the investor has togive notice to the fund about an intention to withdraw money from the fund and the redemptionperiod is the time that the fund takes to return the money after the notice period is over Lockupperiod is the minimum time in quarters that an investor has to wait before withdrawing investedmoney Lockup period is considered to be zero for a fund that has no lock-up provision Subscriptionperiod is the time delay measured in quarters between investing in a fund and actually purchasingfund shares Constrained is an indicator variable that equals one if the fundrsquos strategy belongs toConvertible Arbitrage Fixed Income Arbitrage Emerging Markets and Event Driven strategiesand zero otherwise (Ding et al (2009))
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 50333 798 minus413 034 1014 3701Quarterly returns () 50333 256 minus091 211 566 905AUM ($ million) 50333 2108 110 393 1300 6370Age (quarters) 50333 1756 609 1319 2434 1532Volatility () 50333 533 300 441 804 243Management fees ( annualized) 2998 15 10 15 20 05Incentive fees () 2998 193 200 200 200 34HWM 2998 0694 0000 1000 1000 0461Leverage 2998 0685 0000 1000 1000 0464Open-to-public 2998 0191 0000 0000 0000 0394On-shore 2998 0230 0000 0000 0000 0421Total redemption period (quarters) 2998 1085 0556 0667 1333 1015Lockup period (quarters) 2998 0974 0000 0000 0000 2077Subscription period (quarters) 2998 0356 0333 0333 0333 0190Style
Constrained 2998 0303 0000 0000 1000 0460Constrained = 1
Convertible Arbitrage 2998 0034Emerging Markets 2998 0120Event Driven 2998 0096Fixed Income Arbitrage 2998 0052
Constrained = 0Equity Market Neutral 2998 0073Global Macro 2998 0078LongShort Equity Hedge 2998 0420Multi-Strategy 2998 0077Other 2998 0049
Indirect Incentives of Hedge Fund Managers 879
Our sample period is from January 1995 to December 2010 We focus onthe post-1994 period because TASS started reporting information on ldquodefunctrdquofunds only after 19945 We exclude managed futuresCTAs and funds-of-fundswhich have different incentive fees and are likely to have different inflow-performance relations from typical individual hedge funds We also excludeclosed-end hedge funds as subscriptions in these funds are only possible duringthe initial issuing period so future flows are not possible This initial filterleaves us with 4939 open-end hedge funds6
We drop funds for which TASS does not contain information on organizationalcharacteristics such as management fees incentive fees and HWM provisionsIn addition we consider only funds with an uninterrupted series of net assetvalues and returns so that we can calculate inflows Further we restrict oursample to funds with at least 12 consecutive monthly returns available duringthe sample period If a fund stops reporting returns and then resumes at afuture date we use only the first sequence of uninterrupted data Finally weexclude funds with an incentive fee of zero because there can be no direct payfor performance for these funds
We conduct the analysis using quarterly data because we wish to include onlylagged (not contemporaneous) returns in the flow-performance specificationsand to have a relatively short gap between returns and flows7 To construct aquarterly sample we drop all fund-calendar quarters that have return infor-mation only for a fraction of the period We also require that a sample fundhave a subscription period less than or equal to three months so that quarterlyinflows are not restricted This sample construction process leaves us with asample of 2998 funds (50333 fund-quarter observations)
Table I presents descriptive statistics Time-varying variables such as Quar-terly flow and Quarterly returns are measured at the fund-quarter level andother contractual characteristics such as Management fees and Incentive feesare measured at the fund level8 All time-varying variables except fund age(Age) are winsorized at the 1st and 99th percentiles to minimize the effect ofoutliers
5 Defunct funds include funds that are liquidated merged or restructured as well as those thatstopped reporting returns to TASS (Fung et al (2008))
6 Some funds are closed to new investors but unfortunately we do not know whether a particularfund is taking new money at any point in time so we cannot exclude funds on the basis of thispolicy Including closed funds causes us to understate the flow-performance relation for funds thatare not closed
7 Prior versions of this paper presented estimates using annual data with contemporaneousannual returns included in the flow-performance specifications (Lim Sensoy and Weisbach (2013))The estimates of indirect incentives using annual data are very close to those reported here
8 TASS provides information on fundsrsquo organizational characteristics as of the last available dateof fund data Like most previous studies we also assume that these organizational characteristicsdo not change throughout the life of the fund Agarwal Daniel and Naik (2009) argue that fundsrsquoorganizational characteristics are unlikely to change much over time based on their discussionswith practitioners which suggests that if a manager wants to impose new contractual terms it iseasier for him to start a new fund with different terms than to go through the legal complicationsof changing an existing contract
880 The Journal of Finance Rcopy
Mean Quarterly flow is 80 with a median of 03 so the distributionis highly skewed Mean Quarterly returns is 26 the median is 21 Aver-age fund size (AUM) is $2108 million The remaining variables reflect time-invariant contractual features Summary statistics on these characteristics arevery close to those reported in prior studies (eg Agarwal Daniel and Naik(2009) Baquero and Verbeek (2009) Aragon and Nanda (2012)) The variableManagement fees is the annual percentage of fund assets (AUM) received bythe manager as compensation and has a sample mean (median) of 15 (15)The variable Incentive fees is the percentage of profits above the HWM receivedby the manager as compensation and has a sample mean (median) of 193(20) Over two-thirds of our sample funds 694 have a HWM provision685 report that they use leverage 191 are open to public investors andabout a quarter are on-shore funds
The fee data consist of the fees that are currently publicly quoted by thefunds These data could potentially misrepresent the true fees relevant forour calculations for three reasons First funds sometimes provide fee reduc-tions to particular strategic investors that are not reflected in the database(Ramadorai and Streatfield (2011)) Although we cannot investigate this pos-sibility directly it is unlikely to have a major impact on our conclusions Forexample if the true incentive fee (management fee) averaged over all investorswere 10 lower than what is reported in the database our estimates of indirectincentives using the base GIR model would be overstated by about 1 (5) Asecond issue is that fees can change over time Agarwal and Ray (2012) andDueskar et al (2012) find that fee changes are infrequent and tend to reflectpast performance when they do occur so that fee increases follow good perfor-mance and fee decreases follow poor performance This effect is magnified inindirect incentives because good performance today leads not only to inflowsbut also to higher proportional fees on those inflows Third it is possible thatthere are fee changes unobservable to researchers To the extent that observ-able and unobservable fee changes are positively correlated this possibilityleads us to understate indirect incentives
We consider three variables that reflect potential restrictions on the be-havior of flows First we use Total redemption period defined as the sum ofthe notice period and the redemption period where the notice period is thetime the investor has to give notice to the fund about an intention to with-draw money from the fund and the redemption period is the time that thefund takes to return the money after the notice period is over The Lockupperiod is the minimum time that an investor has to wait before withdrawinginvested money The Subscription period is the time delay between investingin a fund and actually purchasing fund shares The mean total redemption pe-riod lock-up period and subscription period are 109 097 and 036 quartersrespectively
Indirect Incentives of Hedge Fund Managers 881
III Estimating the Sensitivity of Fund Inflows to Performance
To understand the impact of performance on fund flows we employ aBayesian learning framework that presumes investors are continually eval-uating managers trying to assess their abilities (see Berk and Green (2004)Chung et al (2012)) A fundrsquos performance provides information about the man-agerrsquos ability so an observation of performance will lead investors to updatetheir assessment of his ability and allocate more capital to a fund if their es-timate of the managerrsquos ability increases The magnitude of the updates andhence the sensitivity of inflows to performance should depend on the infor-mativeness of the signal relative to the precision of the prior estimate of thefund managerrsquos ability The sensitivity of inflows to performance should alsodepend on the extent to which ability can be scaled to replicate a fundrsquos returndistribution on new capital
Measuring the indirect incentives of hedge fund managers requires an es-timate of the relation between fund performance and future inflows A longliterature beginning with Ippolito (1992) estimates this relation to be rela-tively strong in the mutual fund industry9 It is also positive in the privateequity industry (see eg Kaplan and Schoar (2005)) However the results forhedge funds are less clear Goetzmann Ingersoll and Ross (2003) report a neg-ative and concave relation whereas other studies including Agarwal Danieland Naik (2003) Fung et al (2008) Baquero and Verbeek (2009) and Dinget al (2009) find a positive one
A Empirical Specification
We estimate the following specification
Flowit = β0 +11sumj=1
β0+ j Returnitminus j + γ Xtminus1 + λY + Fixed effects + εit (1)
where Flowit represents flows for fund i in quarter t10
Following the literature on flows to mutual funds (eg Sirri and Tufano(1998) Chevalier and Ellison (1999)) or hedge funds (eg Fung et al (2008)Agarwal Daniel and Naik (2009)) we compute quarterly flows of capital intoa fund as follows
Flowit = AUMit minus AUMitminus1(1 + Returnit
)AUMitminus1
(2)
where AUMit and AUMitminus1 are the AUM of fund i at the end of quarters t andtminus1 respectively and Returnit is the net-of-fee return for fund i in quarter t
9 See Brown Halow and Starks (1996) Sirri and Tufano (1998) Chevalier and Ellison (1999)Barclay Pearson and Weisbach (1998) Del Guercio and Tkac (2002) Bollen (2007) Huang Weiand Yan (2007) and Sensoy (2009)
10 We restrict our estimates to quarterly specifications but prior versions of this paper alsoinclude annual and monthly specifications with similar results to those reported below
882 The Journal of Finance Rcopy
This definition expresses flows as a fraction of beginning-of-period (end ofprior period) AUM which is a natural benchmark from the perspective of afund manager assessing his or her incentives going forward For instance theoption deltas that comprise direct incentives are defined in terms of beginning-of-period AUM An alternative approach to computing fund flows is to scalethe denominator of equation (2) by (1 + Returnit) which expresses flows as afraction of what end-of-period AUM would have been in the absence of flowsAlthough this alternative definition is less intuitive for our purpose it leads tosimilar estimates of indirect incentives
The vector X in equation (1) consists of time-varying fund characteristicsthat include lagged flows the natural logarithm of AUM for fund i at tminus1 thenatural logarithm of one plus fund irsquos age in quarters at tminus1 and fund volatil-ity over the previous 12 months11 The vector Y includes time-invariant fundcharacteristics that include Management fees Incentive fees Total redemptionperiod Lockup period Subscription period and a set of indicator variables thatequal one if the fund has an HWM provision uses leverage is open to publicinvestors and is an on-shore fund All specifications include fixed effects forthe nine styles listed in Table I interacted with calendar quarter fixed effectsThese time-by-style effects capture all shocks observed or unobserved that arecommon to funds of a given style in a given quarter including the returns topeer funds and inflows to other funds of the same style12 Reported standarderrors are robust to heteroskedasticity and account for double clustering byfund and time period13
B Estimates of the Flow-Performance Relation
We present estimates of the flow-performance relation in Panel A ofTable II In Column (1) we include returns in the 11 quarters prior to thecurrent quarter In Column (2) we add a number of fund-level controls In eachspecification the coefficients on returns are positive and statistically signif-icant in most cases and decline sharply over time so the coefficient on themost recent quarterrsquos return is the largest If we sum the coefficients on the 11prior quarters the sum in Column (1) is 144 and in Column (2) is 150 Thesecoefficients imply that a one-standard-deviation increase in returns (9) willlead to about a 13 increase in fund size
11 For young funds that have for example only one yearrsquos worth of return history we cannotcompute lagged returns and flows In such cases we ldquodummy outrdquo missing lagged variables toretain observations To do so we set missing values of lagged flows and returns to zero and includean indicator for missing values
12 Time-by-style fixed effects perform the same adjustment as a factor model regression underthe assumption that the factor loadings are the same for all funds of a given style within a giventime period
13 We focus on linear specifications Although there is some evidence of nonlinear effects in thedata (using splines quadratics etc) they are small in magnitude and have little impact on theestimates obtained with linear specifications
Indirect Incentives of Hedge Fund Managers 883
Table IIEstimates of Flow-Performance Sensitivity
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) using quarterly data The dependent variable is Quarterly flow as defined byequation (2) See Table I for definitions of all independent variables The sample period is from 1995to 2010 We only employ the funds with a subscription period less than or equal to three months toavoid quarterly inflows being restricted by subscription periods Equation (1) is augmented in PanelB to include interaction terms of the log of the fundrsquos age plus one with prior performance and inPanel C to include interaction terms of Constrained with performance variables All specificationsin Panel A and B include style fixed effects quarter fixed effects and quarter-by-style fixed effectsOnly quarter fixed effects are included in Panel C due to multicollinearity Standard errors aredouble clustered by fund and year and indicate statistical significance at the 1 5and 10 level respectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel A Base-case
Quarterly returns(tminus1) 0501 (0045) 0496 (0043)Quarterly returns(tminus2) 0314 (0039) 0263 (0036)Quarterly returns(tminus3) 0209 (0034) 0190 (0028)Quarterly returns(tminus4) 0117 (0035) 0111 (0030)Quarterly returns(tminus5) 0089 (0030) 0098 (0027)Quarterly returns(tminus6) 0054 (0033) 0066 (0030)Quarterly returns(tminus7) 0083 (0027) 0103 (0025)Quarterly returns(tminus8) 0044 (0022) 0060 (0022)Quarterly returns(tminus9) 0012 (0029) 0034 (0029)Quarterly returns(tminus10) minus0010 (0030) 0019 (0029)Quarterly returns(tminus11) 0025 (0027) 0056 (0026)Quarterly flow(tminus1) 0158 (0008)Log(Age + 1) minus0004 (0005)Log(AUM(tminus1)+1) minus0024 (0001)Volatility minus0141 (0037)Management fees 0211 (0386)Incentive fees 0005 (0053)HWM indicator 0020 (0004)Leveraged indicator 0000 (0004)Open-to-public indicator minus0002 (0005)On-shore indicator minus0032 (0005)Log(Total redemption period +1) 0034 (0007)Log(Lock-up period +1) minus0005 (0003)Log(Subscription period +1) minus0023 (0015)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0162 0194F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1940 (0000) 2077 (0000)
(Continued)
884 The Journal of Finance Rcopy
Table IImdashContinued
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel B Age interactions
Quarterly returns(tminus1)timesLog(Age in quarters+1) minus0319 (0035) minus0285 (0034)Quarterly returns(tminus2)timesLog(Age in quarters+1) minus0171 (0027) minus0125 (0025)Quarterly returns(tminus3)timesLog(Age in quarters+1) minus0051 (0036) minus0033 (0036)Quarterly returns(tminus4)timesLog(Age in quarters+1) minus0003 (0032) 0001 (0032)Quarterly returns(tminus5)timesLog(Age in quarters+1) minus0041 (0030) minus0042 (0029)Quarterly returns(tminus6)timesLog(Age in quarters+1) 0038 (0033) 0040 (0031)Quarterly returns(tminus7)timesLog(Age in quarters+1) 0013 (0033) 0007 (0032)Quarterly returns(tminus8)timesLog(Age in quarters+1) 0033 (0033) 0034 (0033)Quarterly returns(tminus9)timesLog(Age in quarters+1) 0003 (0040) minus0004 (0040)Quarterly returns(tminus10)timesLog(Age in quarters+1) 0132 (0038) 0133 (0035)Quarterly returns(tminus11)timesLog(Age in quarters+1) 0018 (0040) 0007 (0041)Quarterly returns(tminus1) 1325 (0117) 1233 (0113)Quarterly returns(tminus2) 0773 (0095) 0602 (0087)Quarterly returns(tminus3) 0354 (0114) 0287 (0110)Quarterly returns(tminus4) 0124 (0112) 0110 (0106)Quarterly returns(tminus5) 0200 (0100) 0215 (0092)Quarterly returns(tminus6) minus0067 (0112) minus0056 (0104)Quarterly returns(tminus7) 0033 (0106) 0079 (0099)Quarterly returns(tminus8) minus0066 (0107) minus0046 (0107)Quarterly returns(tminus9) 0004 (0128) 0053 (0127)Quarterly returns(tminus10) minus0416 (0122) minus0383 (0114)Quarterly returns(tminus11) minus0038 (0126) 0037 (0130)Log(Age + 1) minus0026 (0007) minus0000 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0168 0198F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1788 (0000) 1570 (0000)All age interaction terms 1214 (0000) 1084 (0000)
Panel C Scalability interactions
Quarterly returns(tminus1)timesConstrained minus0215 (0048) minus0215 (0047)Quarterly returns(tminus2)timesConstrained minus0096 (0043) minus0078 (0039)Quarterly returns(tminus3)timesConstrained minus0102 (0038) minus0107 (0037)Quarterly returns(tminus4)timesConstrained 0013 (0034) 0013 (0036)Quarterly returns(tminus5)timesConstrained minus0009 (0032) minus0022 (0032)Quarterly returns(tminus6)timesConstrained 0026 (0034) 0013 (0034)Quarterly returns(tminus7)timesConstrained minus0044 (0036) minus0057 (0036)Quarterly returns(tminus8)timesConstrained 0032 (0037) 0033 (0036)Quarterly returns(tminus9)timesConstrained minus0085 (0030) minus0090 (0031)Quarterly returns(tminus10)timesConstrained minus0023 (0036) minus0007 (0036)Quarterly returns(tminus11)timesConstrained minus0007 (0037) minus0004 (0038)
(Continued)
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 873
The second model that we use augments the GIR model to accommodatefuture performance-based flows The third model is that of Lan Wang andYang (2013 hereafter LWY) in which the manager can endogenously choose theamount of leverage to use at each point in time Finally we present estimatesusing an extension of the LWY model that allows for performance-based flowsas well as endogenous leverage LWY nests GIR which assumes no leverageat any time as a special case so all of our estimates can be thought of asimplications of different versions of the LWY model
Each of these models provides an estimate of the present value of man-agersrsquo compensation per dollar invested in the fund Together with the flow-performance relations these estimates allow us to calculate the magnitude ofindirect incentives facing hedge fund managers For an incremental percent-age point or dollar of current returns to the fundrsquos investors we calculate thepresent value of the additional lifetime income the fundrsquos managers receive inexpectation due to inflows of new investment and the increase in the value ofexisting investorsrsquo assets
As a benchmark for assessing the importance of this indirect pay for perfor-mance we compare its magnitude to the direct performance pay that managersreceive from incentive fees and changes in the value of their own investment inthe fund We use the Agarwal Daniel and Naik (2009) framework to estimatethe change in the value of managersrsquo current compensation (coming from bothincentive fees and the managerrsquos own stake in the fund) for an incrementalreturn
Our estimates indicate that a one-percentage-point increase in returns gen-erates on average $331000 in expected direct incentive pay consisting of$142000 in incremental incentive fees and $190000 in incremental profits onmanagersrsquo personal stakes Using the GIR model with parameter choices thatyield lower-bound estimates we calculate that managers also receive $531000in expected future fee income consisting of $248000 in future fees earned onthe new investment that occurs in response to the incremental performanceand $282000 in extra future fees earned on the increase in the value of exist-ing investorsrsquo assets in the fund Because a one-percentage-point increase inreturn is equal to $211 million for an average-sized fund in our data thesecalculations imply that on average managers receive 16 cents in direct pay andat least 23 cents in indirect pay for each incremental dollar earned for fundinvestors The indirect career-based incentive effect is thus at least 14 timeslarger than the direct income that managers receive from incentive fees andreturns on their personal investments Again this indirect-to-direct incentiveratio of 14 corresponds to model and parameter choices that lead to a lowerbound of estimates of indirect incentives Under other plausible choices theratio would be substantially higher The average indirect-to-direct incentiveratio is 35 across all the models and parameter values we consider
We next find that indirect incentives are even larger for young funds Fornew funds we estimate indirect incentives to be six to 12 times as large asdirect incentives given the parameters used in LWY The importance of indirectincentives declines monotonically as a fund ages largely as a consequence of
874 The Journal of Finance Rcopy
weakening flow-performance relations but continues to be larger than directincentives until the fund is at least 15 years old The importance of indirectincentives also depends on the style of the fund For an average fund followinga style unlikely to be capacity-constrained indirect incentives are 32 to 73times as large as direct incentives while they are 25 to 60 times as large fora fund that is likely to be constrained and hence unable to grow as much inresponse to good performance
Overall our estimates suggest that regardless of the choice of model orreasonable model parameters the total incentives facing hedge fund managersare substantial and much larger than direct incentives alone would implyAlthough direct incentives are themselves substantial indirect incentives inthe hedge fund industry comprise the majority of managersrsquo total incentives
These estimates of substantial indirect incentives in the hedge fund indus-try have a number of implications First they are potentially important forunderstanding hedge fund contracting Hedge fund management contracts arestructured in a sophisticated manner yet perhaps surprisingly we find no ev-idence that direct compensation schemes adjust with the indirect incentivesthat their managers face The lack of such adjustment reflects a larger puzzlein our understanding of markets for alternative assets in that important con-tractual parameters most notably the 20 incentive fee vary little both acrossasset classes and across funds within asset classes
Second institutional investors often state that the financial incentives ofa potential asset manager are an important consideration when deciding be-tween alternative managers Presumably all of a managerrsquos incentives in-cluding both direct and indirect incentives matter in making this choice Theresults discussed above provide estimates of how indirect and total incentivesvary across types of funds and across similar funds of different ages whichshould be relevant for potential investors In addition indirect incentives varysystematically across types of potential investments The results here and inChung et al (2012) provide estimates of indirect and total incentives for hedgefunds and private equity funds Below we also provide estimates of indirectincentives for a sample of 11911 actively managed equity mutual funds overthe period 1995 to 2010 These estimates suggest that indirect incentives inmutual funds are substantial but smaller than those for hedge funds rangingbetween 28 and 66 of the hedge fund estimates
Finally since Fama (1980) and Holmstrom (1982) incentives generated frommanagerial career concerns have been an important part of the theory of thefirm However there are virtually no estimates of their magnitude The esti-mates provided here for hedge fund managers are among the first attempts tomeasure the importance of indirect incentives The fact that career-generatedincentives are so powerful in this industry suggests that they could be equallyimportant in other industries in which they are likely to be harder to estimate
This paper proceeds as follows Section I discusses how we quantify thedirect and indirect components of hedge fund pay for performance SectionII describes the data Section III presents estimates of the flow-performance
Indirect Incentives of Hedge Fund Managers 875
relation Section IV estimates managersrsquo direct and indirect incentives SectionV discusses implications of our estimates Section VI concludes
I Quantifying the Magnitude of Pay for Performance of HedgeFund Managers
A Direct Pay for Performance
Hedge fund managersrsquo compensation generally consists of management feeswhich are a percentage of AUM (often around 15) plus incentive fees whichare a percentage (usually 20) of profits or of profits earned above the HWMIn addition hedge fund managers usually make a personal investment in thefund The direct pay for performance that a manager receives comes from theincentive fees as well as his personal investment in the fund both of whichincrease in value with the fundrsquos performance Quantifying these direct perfor-mance incentives is complicated because of the option-like features containedin the hedge fund managerrsquos incentive fee contract In particular the incen-tive fee contract resembles a portfolio of call options one per investor in thefund The exercise price of each option is determined by the investorrsquos time ofentrance into the fund the fundrsquos hurdle rate and the historical HWM levelpertaining to the investorrsquos assets Thus even if different managers have thesame 20 incentive fee rate their actual direct pay-performance sensitivitywill vary depending on the distance between the current asset value and theexercise prices of these options
To estimate the direct pay-performance sensitivity we use Agarwal Danieland Noikrsquos (2009) total delta approach which measures the impact of an in-cremental one-percentage-point return to fund investors on the value of themanagerrsquos incentive fees plus the increase in the value of the managerrsquos ownownership stake The total delta of the managerrsquos claim to the fund is equalto the sum of these individual optionsrsquo deltas plus the delta of the managerrsquospersonal stake in the fund We follow Agarwal Daniel and Naikrsquos (2009) andassume that the managerrsquos initial stake is zero but the stake grows over timeas managers reinvest all of their incentive fees back into the fund1 Details ofthis calculation are described in Appendix A
B Indirect Pay for Performance
In addition to the pay for performance from incentive fees and their owninvestment in the fund hedge fund managersrsquo lifetime income changes withperformance through a reputational effect good performance increases themarketrsquos perception of a managerrsquos ability leading to higher inflows of newinvestment to the fund Ultimately a fundrsquos managers will receive part of
1 Assuming instead that the managerrsquos initial stake is 1 or 2 of AUM increases the estimatesof indirect incentives by 5 and 10 respectively and the estimates of direct incentives by 2and 5 respectively These changes do not affect our conclusion that indirect incentives are largerelative to direct incentives
876 The Journal of Finance Rcopy
these inflows as future management and incentive fees Furthermore goodperformance mechanically increases the value of existing investorsrsquo stakes inthe fund A portion of this increase will likewise be paid over time in future feesto the fund manager The expectation of this future income depends on todayrsquosperformance leading to what we refer to as indirect incentives
To evaluate the magnitude of these indirect incentives we first need esti-mates of the way in which performance affects expected inflows to the fundThese estimates are discussed in Section III below We also need a model ofthe present value of a managerrsquos expected lifetime compensation as a fractionof fund assets This model should predict for each incremental dollar undermanagement the increase in the managerrsquos expected compensation over thefuture lifetime of the fund We use four such models
The first model is that of GIR The GIR model leads to the lowest estimatesof the sensitivity of future income to todayrsquos performance so it can be thoughtof as providing a lower bound on our estimates of indirect incentives The GIRmodel is a contingent-claims model of the fraction of fund assets that accrue tothe fund manager in expected future compensation Key features of this modelare that compensation is a fixed management fee plus a percentage of profitsabove a HWM the fundrsquos asset value follows a martingale with drift generatedby the managerrsquos alpha investors continuously withdraw assets (eg an en-dowment investor might withdraw 5 per year) and an investor liquidates hisor her position following a sufficiently negative return shock (so that expectedfund life is finite)
The GIR model does not account for all factors that could affect managersrsquofuture compensation and in turn indirect incentives In particular the GIRmodel does not allow the fundrsquos asset value to grow through future performance-based fund flows Under appropriate assumptions the effect of such flows inthe context of the GIR model is to increase the variance of the fundrsquos AUM2
For this reason our second model reports estimates from a variant of the GIRmodel in which fund variance is augmented
Another factor not accounted for by the GIR model is the fact that a fundrsquosportfolio is endogenously chosen with managers adjusting their portfolios tomaximize their incomes given a particular incentive scheme Following LWYour third model allows managers to perform such adjustments by levering theirfunds strategically Given that hedge fund managers typically do use leverageestimates of indirect incentives that incorporate this feature are likely to bemore accurate
Finally we present estimates from a fourth model also introduced by LWYthat augments the basic LWY model to include future performance-based fundflows Note also that the LWY model nests the GIR model as a special case in
2 Suppose as an approximation that the flow-performance relationship is linear in logs andflows are contemporaneous with returns Suppose that without flows the log AUM of the fundevolves as a martingale as in the GIR model st+1 = st + et+1 With performance-based flows st+1 =st + get+1 where g gt 1 captures the flow effect Therefore even with performance-based flows thelog AUM would still follow a martingale so the GIR model still applies but with a higher variancethan in the case of no flows We thank the Associate Editor for suggesting this argument to us
Indirect Incentives of Hedge Fund Managers 877
which leverage is zero so all four sets of estimates can be viewed as comingfrom variations of the LWY model
Using different models to estimate indirect incentives allows us to gaugethe importance of different factors especially the importance of futureperformance-based fund flows and endogenous portfolio choice in determin-ing the fraction of a fundrsquos value that will ultimately accrue to managers infuture fees3 Details on all four models our choices of model parameters andthe associated calculations are described in Appendix B
For each of the models above we estimate the present value of the total(management plus incentive) future fees that the manager earns on an extradollar of AUM To calculate indirect pay for performance we multiply thispresent value by an estimate of the number of incremental dollars of AUMthat result from a one-percentage-point increase in returns to investors Thelatter increase consists of two parts namely the mechanical increase in thevalue of existing investorsrsquo stakes plus incremental inflows of new investmentIn this way the models for a managerrsquos fee value together with our estimatesof the flow-performance relation facing hedge fund managers provide an esti-mate of the present value of the incremental future revenue that the hedgefund manager expects to earn as a result of a one-percentage-point improve-ment in current net returns4 This calculation results in an incentive measuredenominated in dollars per percentage point change in returns We also con-duct the same calculation for a one-dollar improvement in investor return toobtain a measure in terms of dollars in managerial wealth per dollar returnedto shareholders Both of these measures are commonly used in the literatureon manager incentives
Because the GIR and LWY models estimate present values no further ad-justment for the riskiness of future income is required Also the estimates donot require that the manager continue to manage the fund in the future underthe assumption that the present value of the managerrsquos claims to future feeincome can be monetized when the manager departs
II Hedge Fund Data
Our data come from the TASS database which covers about 40 of thehedge fund universe (Agarwal Daniel and Naik (2009)) Summary statisticsfor key sample fund characteristics reported in Table I are very close to thosefor the sample considered by Agarwal Daniel and Naik (2009) who mergeand consolidate four major databases (CISDM HFR MSCI and TASS) Wetherefore conclude that our sample of hedge funds is representative of thehedge fund universe
3 In addition to GIR and LWY there have been a number of other attempts to value managersrsquoclaims to hedge funds Important contributions include Panageas and Westerfield (2009) Drechsler(2014) and Guasoni and Oblόj (2013)
4 Net returns can be improved either through increased gross returns or decreased costs borneby the fund such as financing costs security lending fees and settlement charges The incentiveswe measure are therefore incentives to achieve both
878 The Journal of Finance Rcopy
Table ISummary Statistics
This table presents descriptive statistics for the sample of 2998 funds during the 1995 to 2010sample period Time-varying variables are reported at the fund-quarter level and other time-invariant contractual characteristics are reported at the fund level Quarterly flow is the differencebetween the reported quarter-end AUM and the quarter-beginning AUM times (1 + Quarterlyreturns) scaled by quarter-beginning AUM Quarterly returns are the reported quarterly net-of-fee returns AUM is assets under management in millions measured at the end of each quarterAge is the number of quarters from the fundrsquos inception date measured at the beginning of eachquarter Volatility is the standard deviation of monthly (net-of-fee) returns over the 12 monthsprior to each quarter multiplied by the square root of three to arrive at a quarterly figure Alltime-varying variables except Age are winsorized at the 1st and 99th percentiles Management feesare the percentage of the assets charged by the fund as regular fees in annual terms Incentive feesare the percentage of positive profits received by the manager as performance incentives HWMis an indicator variable that equals one if the fund has a HWM provision and zero otherwiseLeverage is an indicator variable that equals one if the fund uses leverage and zero otherwiseOpen-to-public is an indicator variable that equals one if the fund allows public investment andzero otherwise On-shore is an indicator variable that equals one if the fund is domiciled in theUnited States and zero otherwise Total redemption period is the sum of the notice period and theredemption period measured in quarters where the notice period is the time the investor has togive notice to the fund about an intention to withdraw money from the fund and the redemptionperiod is the time that the fund takes to return the money after the notice period is over Lockupperiod is the minimum time in quarters that an investor has to wait before withdrawing investedmoney Lockup period is considered to be zero for a fund that has no lock-up provision Subscriptionperiod is the time delay measured in quarters between investing in a fund and actually purchasingfund shares Constrained is an indicator variable that equals one if the fundrsquos strategy belongs toConvertible Arbitrage Fixed Income Arbitrage Emerging Markets and Event Driven strategiesand zero otherwise (Ding et al (2009))
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 50333 798 minus413 034 1014 3701Quarterly returns () 50333 256 minus091 211 566 905AUM ($ million) 50333 2108 110 393 1300 6370Age (quarters) 50333 1756 609 1319 2434 1532Volatility () 50333 533 300 441 804 243Management fees ( annualized) 2998 15 10 15 20 05Incentive fees () 2998 193 200 200 200 34HWM 2998 0694 0000 1000 1000 0461Leverage 2998 0685 0000 1000 1000 0464Open-to-public 2998 0191 0000 0000 0000 0394On-shore 2998 0230 0000 0000 0000 0421Total redemption period (quarters) 2998 1085 0556 0667 1333 1015Lockup period (quarters) 2998 0974 0000 0000 0000 2077Subscription period (quarters) 2998 0356 0333 0333 0333 0190Style
Constrained 2998 0303 0000 0000 1000 0460Constrained = 1
Convertible Arbitrage 2998 0034Emerging Markets 2998 0120Event Driven 2998 0096Fixed Income Arbitrage 2998 0052
Constrained = 0Equity Market Neutral 2998 0073Global Macro 2998 0078LongShort Equity Hedge 2998 0420Multi-Strategy 2998 0077Other 2998 0049
Indirect Incentives of Hedge Fund Managers 879
Our sample period is from January 1995 to December 2010 We focus onthe post-1994 period because TASS started reporting information on ldquodefunctrdquofunds only after 19945 We exclude managed futuresCTAs and funds-of-fundswhich have different incentive fees and are likely to have different inflow-performance relations from typical individual hedge funds We also excludeclosed-end hedge funds as subscriptions in these funds are only possible duringthe initial issuing period so future flows are not possible This initial filterleaves us with 4939 open-end hedge funds6
We drop funds for which TASS does not contain information on organizationalcharacteristics such as management fees incentive fees and HWM provisionsIn addition we consider only funds with an uninterrupted series of net assetvalues and returns so that we can calculate inflows Further we restrict oursample to funds with at least 12 consecutive monthly returns available duringthe sample period If a fund stops reporting returns and then resumes at afuture date we use only the first sequence of uninterrupted data Finally weexclude funds with an incentive fee of zero because there can be no direct payfor performance for these funds
We conduct the analysis using quarterly data because we wish to include onlylagged (not contemporaneous) returns in the flow-performance specificationsand to have a relatively short gap between returns and flows7 To construct aquarterly sample we drop all fund-calendar quarters that have return infor-mation only for a fraction of the period We also require that a sample fundhave a subscription period less than or equal to three months so that quarterlyinflows are not restricted This sample construction process leaves us with asample of 2998 funds (50333 fund-quarter observations)
Table I presents descriptive statistics Time-varying variables such as Quar-terly flow and Quarterly returns are measured at the fund-quarter level andother contractual characteristics such as Management fees and Incentive feesare measured at the fund level8 All time-varying variables except fund age(Age) are winsorized at the 1st and 99th percentiles to minimize the effect ofoutliers
5 Defunct funds include funds that are liquidated merged or restructured as well as those thatstopped reporting returns to TASS (Fung et al (2008))
6 Some funds are closed to new investors but unfortunately we do not know whether a particularfund is taking new money at any point in time so we cannot exclude funds on the basis of thispolicy Including closed funds causes us to understate the flow-performance relation for funds thatare not closed
7 Prior versions of this paper presented estimates using annual data with contemporaneousannual returns included in the flow-performance specifications (Lim Sensoy and Weisbach (2013))The estimates of indirect incentives using annual data are very close to those reported here
8 TASS provides information on fundsrsquo organizational characteristics as of the last available dateof fund data Like most previous studies we also assume that these organizational characteristicsdo not change throughout the life of the fund Agarwal Daniel and Naik (2009) argue that fundsrsquoorganizational characteristics are unlikely to change much over time based on their discussionswith practitioners which suggests that if a manager wants to impose new contractual terms it iseasier for him to start a new fund with different terms than to go through the legal complicationsof changing an existing contract
880 The Journal of Finance Rcopy
Mean Quarterly flow is 80 with a median of 03 so the distributionis highly skewed Mean Quarterly returns is 26 the median is 21 Aver-age fund size (AUM) is $2108 million The remaining variables reflect time-invariant contractual features Summary statistics on these characteristics arevery close to those reported in prior studies (eg Agarwal Daniel and Naik(2009) Baquero and Verbeek (2009) Aragon and Nanda (2012)) The variableManagement fees is the annual percentage of fund assets (AUM) received bythe manager as compensation and has a sample mean (median) of 15 (15)The variable Incentive fees is the percentage of profits above the HWM receivedby the manager as compensation and has a sample mean (median) of 193(20) Over two-thirds of our sample funds 694 have a HWM provision685 report that they use leverage 191 are open to public investors andabout a quarter are on-shore funds
The fee data consist of the fees that are currently publicly quoted by thefunds These data could potentially misrepresent the true fees relevant forour calculations for three reasons First funds sometimes provide fee reduc-tions to particular strategic investors that are not reflected in the database(Ramadorai and Streatfield (2011)) Although we cannot investigate this pos-sibility directly it is unlikely to have a major impact on our conclusions Forexample if the true incentive fee (management fee) averaged over all investorswere 10 lower than what is reported in the database our estimates of indirectincentives using the base GIR model would be overstated by about 1 (5) Asecond issue is that fees can change over time Agarwal and Ray (2012) andDueskar et al (2012) find that fee changes are infrequent and tend to reflectpast performance when they do occur so that fee increases follow good perfor-mance and fee decreases follow poor performance This effect is magnified inindirect incentives because good performance today leads not only to inflowsbut also to higher proportional fees on those inflows Third it is possible thatthere are fee changes unobservable to researchers To the extent that observ-able and unobservable fee changes are positively correlated this possibilityleads us to understate indirect incentives
We consider three variables that reflect potential restrictions on the be-havior of flows First we use Total redemption period defined as the sum ofthe notice period and the redemption period where the notice period is thetime the investor has to give notice to the fund about an intention to with-draw money from the fund and the redemption period is the time that thefund takes to return the money after the notice period is over The Lockupperiod is the minimum time that an investor has to wait before withdrawinginvested money The Subscription period is the time delay between investingin a fund and actually purchasing fund shares The mean total redemption pe-riod lock-up period and subscription period are 109 097 and 036 quartersrespectively
Indirect Incentives of Hedge Fund Managers 881
III Estimating the Sensitivity of Fund Inflows to Performance
To understand the impact of performance on fund flows we employ aBayesian learning framework that presumes investors are continually eval-uating managers trying to assess their abilities (see Berk and Green (2004)Chung et al (2012)) A fundrsquos performance provides information about the man-agerrsquos ability so an observation of performance will lead investors to updatetheir assessment of his ability and allocate more capital to a fund if their es-timate of the managerrsquos ability increases The magnitude of the updates andhence the sensitivity of inflows to performance should depend on the infor-mativeness of the signal relative to the precision of the prior estimate of thefund managerrsquos ability The sensitivity of inflows to performance should alsodepend on the extent to which ability can be scaled to replicate a fundrsquos returndistribution on new capital
Measuring the indirect incentives of hedge fund managers requires an es-timate of the relation between fund performance and future inflows A longliterature beginning with Ippolito (1992) estimates this relation to be rela-tively strong in the mutual fund industry9 It is also positive in the privateequity industry (see eg Kaplan and Schoar (2005)) However the results forhedge funds are less clear Goetzmann Ingersoll and Ross (2003) report a neg-ative and concave relation whereas other studies including Agarwal Danieland Naik (2003) Fung et al (2008) Baquero and Verbeek (2009) and Dinget al (2009) find a positive one
A Empirical Specification
We estimate the following specification
Flowit = β0 +11sumj=1
β0+ j Returnitminus j + γ Xtminus1 + λY + Fixed effects + εit (1)
where Flowit represents flows for fund i in quarter t10
Following the literature on flows to mutual funds (eg Sirri and Tufano(1998) Chevalier and Ellison (1999)) or hedge funds (eg Fung et al (2008)Agarwal Daniel and Naik (2009)) we compute quarterly flows of capital intoa fund as follows
Flowit = AUMit minus AUMitminus1(1 + Returnit
)AUMitminus1
(2)
where AUMit and AUMitminus1 are the AUM of fund i at the end of quarters t andtminus1 respectively and Returnit is the net-of-fee return for fund i in quarter t
9 See Brown Halow and Starks (1996) Sirri and Tufano (1998) Chevalier and Ellison (1999)Barclay Pearson and Weisbach (1998) Del Guercio and Tkac (2002) Bollen (2007) Huang Weiand Yan (2007) and Sensoy (2009)
10 We restrict our estimates to quarterly specifications but prior versions of this paper alsoinclude annual and monthly specifications with similar results to those reported below
882 The Journal of Finance Rcopy
This definition expresses flows as a fraction of beginning-of-period (end ofprior period) AUM which is a natural benchmark from the perspective of afund manager assessing his or her incentives going forward For instance theoption deltas that comprise direct incentives are defined in terms of beginning-of-period AUM An alternative approach to computing fund flows is to scalethe denominator of equation (2) by (1 + Returnit) which expresses flows as afraction of what end-of-period AUM would have been in the absence of flowsAlthough this alternative definition is less intuitive for our purpose it leads tosimilar estimates of indirect incentives
The vector X in equation (1) consists of time-varying fund characteristicsthat include lagged flows the natural logarithm of AUM for fund i at tminus1 thenatural logarithm of one plus fund irsquos age in quarters at tminus1 and fund volatil-ity over the previous 12 months11 The vector Y includes time-invariant fundcharacteristics that include Management fees Incentive fees Total redemptionperiod Lockup period Subscription period and a set of indicator variables thatequal one if the fund has an HWM provision uses leverage is open to publicinvestors and is an on-shore fund All specifications include fixed effects forthe nine styles listed in Table I interacted with calendar quarter fixed effectsThese time-by-style effects capture all shocks observed or unobserved that arecommon to funds of a given style in a given quarter including the returns topeer funds and inflows to other funds of the same style12 Reported standarderrors are robust to heteroskedasticity and account for double clustering byfund and time period13
B Estimates of the Flow-Performance Relation
We present estimates of the flow-performance relation in Panel A ofTable II In Column (1) we include returns in the 11 quarters prior to thecurrent quarter In Column (2) we add a number of fund-level controls In eachspecification the coefficients on returns are positive and statistically signif-icant in most cases and decline sharply over time so the coefficient on themost recent quarterrsquos return is the largest If we sum the coefficients on the 11prior quarters the sum in Column (1) is 144 and in Column (2) is 150 Thesecoefficients imply that a one-standard-deviation increase in returns (9) willlead to about a 13 increase in fund size
11 For young funds that have for example only one yearrsquos worth of return history we cannotcompute lagged returns and flows In such cases we ldquodummy outrdquo missing lagged variables toretain observations To do so we set missing values of lagged flows and returns to zero and includean indicator for missing values
12 Time-by-style fixed effects perform the same adjustment as a factor model regression underthe assumption that the factor loadings are the same for all funds of a given style within a giventime period
13 We focus on linear specifications Although there is some evidence of nonlinear effects in thedata (using splines quadratics etc) they are small in magnitude and have little impact on theestimates obtained with linear specifications
Indirect Incentives of Hedge Fund Managers 883
Table IIEstimates of Flow-Performance Sensitivity
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) using quarterly data The dependent variable is Quarterly flow as defined byequation (2) See Table I for definitions of all independent variables The sample period is from 1995to 2010 We only employ the funds with a subscription period less than or equal to three months toavoid quarterly inflows being restricted by subscription periods Equation (1) is augmented in PanelB to include interaction terms of the log of the fundrsquos age plus one with prior performance and inPanel C to include interaction terms of Constrained with performance variables All specificationsin Panel A and B include style fixed effects quarter fixed effects and quarter-by-style fixed effectsOnly quarter fixed effects are included in Panel C due to multicollinearity Standard errors aredouble clustered by fund and year and indicate statistical significance at the 1 5and 10 level respectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel A Base-case
Quarterly returns(tminus1) 0501 (0045) 0496 (0043)Quarterly returns(tminus2) 0314 (0039) 0263 (0036)Quarterly returns(tminus3) 0209 (0034) 0190 (0028)Quarterly returns(tminus4) 0117 (0035) 0111 (0030)Quarterly returns(tminus5) 0089 (0030) 0098 (0027)Quarterly returns(tminus6) 0054 (0033) 0066 (0030)Quarterly returns(tminus7) 0083 (0027) 0103 (0025)Quarterly returns(tminus8) 0044 (0022) 0060 (0022)Quarterly returns(tminus9) 0012 (0029) 0034 (0029)Quarterly returns(tminus10) minus0010 (0030) 0019 (0029)Quarterly returns(tminus11) 0025 (0027) 0056 (0026)Quarterly flow(tminus1) 0158 (0008)Log(Age + 1) minus0004 (0005)Log(AUM(tminus1)+1) minus0024 (0001)Volatility minus0141 (0037)Management fees 0211 (0386)Incentive fees 0005 (0053)HWM indicator 0020 (0004)Leveraged indicator 0000 (0004)Open-to-public indicator minus0002 (0005)On-shore indicator minus0032 (0005)Log(Total redemption period +1) 0034 (0007)Log(Lock-up period +1) minus0005 (0003)Log(Subscription period +1) minus0023 (0015)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0162 0194F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1940 (0000) 2077 (0000)
(Continued)
884 The Journal of Finance Rcopy
Table IImdashContinued
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel B Age interactions
Quarterly returns(tminus1)timesLog(Age in quarters+1) minus0319 (0035) minus0285 (0034)Quarterly returns(tminus2)timesLog(Age in quarters+1) minus0171 (0027) minus0125 (0025)Quarterly returns(tminus3)timesLog(Age in quarters+1) minus0051 (0036) minus0033 (0036)Quarterly returns(tminus4)timesLog(Age in quarters+1) minus0003 (0032) 0001 (0032)Quarterly returns(tminus5)timesLog(Age in quarters+1) minus0041 (0030) minus0042 (0029)Quarterly returns(tminus6)timesLog(Age in quarters+1) 0038 (0033) 0040 (0031)Quarterly returns(tminus7)timesLog(Age in quarters+1) 0013 (0033) 0007 (0032)Quarterly returns(tminus8)timesLog(Age in quarters+1) 0033 (0033) 0034 (0033)Quarterly returns(tminus9)timesLog(Age in quarters+1) 0003 (0040) minus0004 (0040)Quarterly returns(tminus10)timesLog(Age in quarters+1) 0132 (0038) 0133 (0035)Quarterly returns(tminus11)timesLog(Age in quarters+1) 0018 (0040) 0007 (0041)Quarterly returns(tminus1) 1325 (0117) 1233 (0113)Quarterly returns(tminus2) 0773 (0095) 0602 (0087)Quarterly returns(tminus3) 0354 (0114) 0287 (0110)Quarterly returns(tminus4) 0124 (0112) 0110 (0106)Quarterly returns(tminus5) 0200 (0100) 0215 (0092)Quarterly returns(tminus6) minus0067 (0112) minus0056 (0104)Quarterly returns(tminus7) 0033 (0106) 0079 (0099)Quarterly returns(tminus8) minus0066 (0107) minus0046 (0107)Quarterly returns(tminus9) 0004 (0128) 0053 (0127)Quarterly returns(tminus10) minus0416 (0122) minus0383 (0114)Quarterly returns(tminus11) minus0038 (0126) 0037 (0130)Log(Age + 1) minus0026 (0007) minus0000 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0168 0198F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1788 (0000) 1570 (0000)All age interaction terms 1214 (0000) 1084 (0000)
Panel C Scalability interactions
Quarterly returns(tminus1)timesConstrained minus0215 (0048) minus0215 (0047)Quarterly returns(tminus2)timesConstrained minus0096 (0043) minus0078 (0039)Quarterly returns(tminus3)timesConstrained minus0102 (0038) minus0107 (0037)Quarterly returns(tminus4)timesConstrained 0013 (0034) 0013 (0036)Quarterly returns(tminus5)timesConstrained minus0009 (0032) minus0022 (0032)Quarterly returns(tminus6)timesConstrained 0026 (0034) 0013 (0034)Quarterly returns(tminus7)timesConstrained minus0044 (0036) minus0057 (0036)Quarterly returns(tminus8)timesConstrained 0032 (0037) 0033 (0036)Quarterly returns(tminus9)timesConstrained minus0085 (0030) minus0090 (0031)Quarterly returns(tminus10)timesConstrained minus0023 (0036) minus0007 (0036)Quarterly returns(tminus11)timesConstrained minus0007 (0037) minus0004 (0038)
(Continued)
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
874 The Journal of Finance Rcopy
weakening flow-performance relations but continues to be larger than directincentives until the fund is at least 15 years old The importance of indirectincentives also depends on the style of the fund For an average fund followinga style unlikely to be capacity-constrained indirect incentives are 32 to 73times as large as direct incentives while they are 25 to 60 times as large fora fund that is likely to be constrained and hence unable to grow as much inresponse to good performance
Overall our estimates suggest that regardless of the choice of model orreasonable model parameters the total incentives facing hedge fund managersare substantial and much larger than direct incentives alone would implyAlthough direct incentives are themselves substantial indirect incentives inthe hedge fund industry comprise the majority of managersrsquo total incentives
These estimates of substantial indirect incentives in the hedge fund indus-try have a number of implications First they are potentially important forunderstanding hedge fund contracting Hedge fund management contracts arestructured in a sophisticated manner yet perhaps surprisingly we find no ev-idence that direct compensation schemes adjust with the indirect incentivesthat their managers face The lack of such adjustment reflects a larger puzzlein our understanding of markets for alternative assets in that important con-tractual parameters most notably the 20 incentive fee vary little both acrossasset classes and across funds within asset classes
Second institutional investors often state that the financial incentives ofa potential asset manager are an important consideration when deciding be-tween alternative managers Presumably all of a managerrsquos incentives in-cluding both direct and indirect incentives matter in making this choice Theresults discussed above provide estimates of how indirect and total incentivesvary across types of funds and across similar funds of different ages whichshould be relevant for potential investors In addition indirect incentives varysystematically across types of potential investments The results here and inChung et al (2012) provide estimates of indirect and total incentives for hedgefunds and private equity funds Below we also provide estimates of indirectincentives for a sample of 11911 actively managed equity mutual funds overthe period 1995 to 2010 These estimates suggest that indirect incentives inmutual funds are substantial but smaller than those for hedge funds rangingbetween 28 and 66 of the hedge fund estimates
Finally since Fama (1980) and Holmstrom (1982) incentives generated frommanagerial career concerns have been an important part of the theory of thefirm However there are virtually no estimates of their magnitude The esti-mates provided here for hedge fund managers are among the first attempts tomeasure the importance of indirect incentives The fact that career-generatedincentives are so powerful in this industry suggests that they could be equallyimportant in other industries in which they are likely to be harder to estimate
This paper proceeds as follows Section I discusses how we quantify thedirect and indirect components of hedge fund pay for performance SectionII describes the data Section III presents estimates of the flow-performance
Indirect Incentives of Hedge Fund Managers 875
relation Section IV estimates managersrsquo direct and indirect incentives SectionV discusses implications of our estimates Section VI concludes
I Quantifying the Magnitude of Pay for Performance of HedgeFund Managers
A Direct Pay for Performance
Hedge fund managersrsquo compensation generally consists of management feeswhich are a percentage of AUM (often around 15) plus incentive fees whichare a percentage (usually 20) of profits or of profits earned above the HWMIn addition hedge fund managers usually make a personal investment in thefund The direct pay for performance that a manager receives comes from theincentive fees as well as his personal investment in the fund both of whichincrease in value with the fundrsquos performance Quantifying these direct perfor-mance incentives is complicated because of the option-like features containedin the hedge fund managerrsquos incentive fee contract In particular the incen-tive fee contract resembles a portfolio of call options one per investor in thefund The exercise price of each option is determined by the investorrsquos time ofentrance into the fund the fundrsquos hurdle rate and the historical HWM levelpertaining to the investorrsquos assets Thus even if different managers have thesame 20 incentive fee rate their actual direct pay-performance sensitivitywill vary depending on the distance between the current asset value and theexercise prices of these options
To estimate the direct pay-performance sensitivity we use Agarwal Danieland Noikrsquos (2009) total delta approach which measures the impact of an in-cremental one-percentage-point return to fund investors on the value of themanagerrsquos incentive fees plus the increase in the value of the managerrsquos ownownership stake The total delta of the managerrsquos claim to the fund is equalto the sum of these individual optionsrsquo deltas plus the delta of the managerrsquospersonal stake in the fund We follow Agarwal Daniel and Naikrsquos (2009) andassume that the managerrsquos initial stake is zero but the stake grows over timeas managers reinvest all of their incentive fees back into the fund1 Details ofthis calculation are described in Appendix A
B Indirect Pay for Performance
In addition to the pay for performance from incentive fees and their owninvestment in the fund hedge fund managersrsquo lifetime income changes withperformance through a reputational effect good performance increases themarketrsquos perception of a managerrsquos ability leading to higher inflows of newinvestment to the fund Ultimately a fundrsquos managers will receive part of
1 Assuming instead that the managerrsquos initial stake is 1 or 2 of AUM increases the estimatesof indirect incentives by 5 and 10 respectively and the estimates of direct incentives by 2and 5 respectively These changes do not affect our conclusion that indirect incentives are largerelative to direct incentives
876 The Journal of Finance Rcopy
these inflows as future management and incentive fees Furthermore goodperformance mechanically increases the value of existing investorsrsquo stakes inthe fund A portion of this increase will likewise be paid over time in future feesto the fund manager The expectation of this future income depends on todayrsquosperformance leading to what we refer to as indirect incentives
To evaluate the magnitude of these indirect incentives we first need esti-mates of the way in which performance affects expected inflows to the fundThese estimates are discussed in Section III below We also need a model ofthe present value of a managerrsquos expected lifetime compensation as a fractionof fund assets This model should predict for each incremental dollar undermanagement the increase in the managerrsquos expected compensation over thefuture lifetime of the fund We use four such models
The first model is that of GIR The GIR model leads to the lowest estimatesof the sensitivity of future income to todayrsquos performance so it can be thoughtof as providing a lower bound on our estimates of indirect incentives The GIRmodel is a contingent-claims model of the fraction of fund assets that accrue tothe fund manager in expected future compensation Key features of this modelare that compensation is a fixed management fee plus a percentage of profitsabove a HWM the fundrsquos asset value follows a martingale with drift generatedby the managerrsquos alpha investors continuously withdraw assets (eg an en-dowment investor might withdraw 5 per year) and an investor liquidates hisor her position following a sufficiently negative return shock (so that expectedfund life is finite)
The GIR model does not account for all factors that could affect managersrsquofuture compensation and in turn indirect incentives In particular the GIRmodel does not allow the fundrsquos asset value to grow through future performance-based fund flows Under appropriate assumptions the effect of such flows inthe context of the GIR model is to increase the variance of the fundrsquos AUM2
For this reason our second model reports estimates from a variant of the GIRmodel in which fund variance is augmented
Another factor not accounted for by the GIR model is the fact that a fundrsquosportfolio is endogenously chosen with managers adjusting their portfolios tomaximize their incomes given a particular incentive scheme Following LWYour third model allows managers to perform such adjustments by levering theirfunds strategically Given that hedge fund managers typically do use leverageestimates of indirect incentives that incorporate this feature are likely to bemore accurate
Finally we present estimates from a fourth model also introduced by LWYthat augments the basic LWY model to include future performance-based fundflows Note also that the LWY model nests the GIR model as a special case in
2 Suppose as an approximation that the flow-performance relationship is linear in logs andflows are contemporaneous with returns Suppose that without flows the log AUM of the fundevolves as a martingale as in the GIR model st+1 = st + et+1 With performance-based flows st+1 =st + get+1 where g gt 1 captures the flow effect Therefore even with performance-based flows thelog AUM would still follow a martingale so the GIR model still applies but with a higher variancethan in the case of no flows We thank the Associate Editor for suggesting this argument to us
Indirect Incentives of Hedge Fund Managers 877
which leverage is zero so all four sets of estimates can be viewed as comingfrom variations of the LWY model
Using different models to estimate indirect incentives allows us to gaugethe importance of different factors especially the importance of futureperformance-based fund flows and endogenous portfolio choice in determin-ing the fraction of a fundrsquos value that will ultimately accrue to managers infuture fees3 Details on all four models our choices of model parameters andthe associated calculations are described in Appendix B
For each of the models above we estimate the present value of the total(management plus incentive) future fees that the manager earns on an extradollar of AUM To calculate indirect pay for performance we multiply thispresent value by an estimate of the number of incremental dollars of AUMthat result from a one-percentage-point increase in returns to investors Thelatter increase consists of two parts namely the mechanical increase in thevalue of existing investorsrsquo stakes plus incremental inflows of new investmentIn this way the models for a managerrsquos fee value together with our estimatesof the flow-performance relation facing hedge fund managers provide an esti-mate of the present value of the incremental future revenue that the hedgefund manager expects to earn as a result of a one-percentage-point improve-ment in current net returns4 This calculation results in an incentive measuredenominated in dollars per percentage point change in returns We also con-duct the same calculation for a one-dollar improvement in investor return toobtain a measure in terms of dollars in managerial wealth per dollar returnedto shareholders Both of these measures are commonly used in the literatureon manager incentives
Because the GIR and LWY models estimate present values no further ad-justment for the riskiness of future income is required Also the estimates donot require that the manager continue to manage the fund in the future underthe assumption that the present value of the managerrsquos claims to future feeincome can be monetized when the manager departs
II Hedge Fund Data
Our data come from the TASS database which covers about 40 of thehedge fund universe (Agarwal Daniel and Naik (2009)) Summary statisticsfor key sample fund characteristics reported in Table I are very close to thosefor the sample considered by Agarwal Daniel and Naik (2009) who mergeand consolidate four major databases (CISDM HFR MSCI and TASS) Wetherefore conclude that our sample of hedge funds is representative of thehedge fund universe
3 In addition to GIR and LWY there have been a number of other attempts to value managersrsquoclaims to hedge funds Important contributions include Panageas and Westerfield (2009) Drechsler(2014) and Guasoni and Oblόj (2013)
4 Net returns can be improved either through increased gross returns or decreased costs borneby the fund such as financing costs security lending fees and settlement charges The incentiveswe measure are therefore incentives to achieve both
878 The Journal of Finance Rcopy
Table ISummary Statistics
This table presents descriptive statistics for the sample of 2998 funds during the 1995 to 2010sample period Time-varying variables are reported at the fund-quarter level and other time-invariant contractual characteristics are reported at the fund level Quarterly flow is the differencebetween the reported quarter-end AUM and the quarter-beginning AUM times (1 + Quarterlyreturns) scaled by quarter-beginning AUM Quarterly returns are the reported quarterly net-of-fee returns AUM is assets under management in millions measured at the end of each quarterAge is the number of quarters from the fundrsquos inception date measured at the beginning of eachquarter Volatility is the standard deviation of monthly (net-of-fee) returns over the 12 monthsprior to each quarter multiplied by the square root of three to arrive at a quarterly figure Alltime-varying variables except Age are winsorized at the 1st and 99th percentiles Management feesare the percentage of the assets charged by the fund as regular fees in annual terms Incentive feesare the percentage of positive profits received by the manager as performance incentives HWMis an indicator variable that equals one if the fund has a HWM provision and zero otherwiseLeverage is an indicator variable that equals one if the fund uses leverage and zero otherwiseOpen-to-public is an indicator variable that equals one if the fund allows public investment andzero otherwise On-shore is an indicator variable that equals one if the fund is domiciled in theUnited States and zero otherwise Total redemption period is the sum of the notice period and theredemption period measured in quarters where the notice period is the time the investor has togive notice to the fund about an intention to withdraw money from the fund and the redemptionperiod is the time that the fund takes to return the money after the notice period is over Lockupperiod is the minimum time in quarters that an investor has to wait before withdrawing investedmoney Lockup period is considered to be zero for a fund that has no lock-up provision Subscriptionperiod is the time delay measured in quarters between investing in a fund and actually purchasingfund shares Constrained is an indicator variable that equals one if the fundrsquos strategy belongs toConvertible Arbitrage Fixed Income Arbitrage Emerging Markets and Event Driven strategiesand zero otherwise (Ding et al (2009))
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 50333 798 minus413 034 1014 3701Quarterly returns () 50333 256 minus091 211 566 905AUM ($ million) 50333 2108 110 393 1300 6370Age (quarters) 50333 1756 609 1319 2434 1532Volatility () 50333 533 300 441 804 243Management fees ( annualized) 2998 15 10 15 20 05Incentive fees () 2998 193 200 200 200 34HWM 2998 0694 0000 1000 1000 0461Leverage 2998 0685 0000 1000 1000 0464Open-to-public 2998 0191 0000 0000 0000 0394On-shore 2998 0230 0000 0000 0000 0421Total redemption period (quarters) 2998 1085 0556 0667 1333 1015Lockup period (quarters) 2998 0974 0000 0000 0000 2077Subscription period (quarters) 2998 0356 0333 0333 0333 0190Style
Constrained 2998 0303 0000 0000 1000 0460Constrained = 1
Convertible Arbitrage 2998 0034Emerging Markets 2998 0120Event Driven 2998 0096Fixed Income Arbitrage 2998 0052
Constrained = 0Equity Market Neutral 2998 0073Global Macro 2998 0078LongShort Equity Hedge 2998 0420Multi-Strategy 2998 0077Other 2998 0049
Indirect Incentives of Hedge Fund Managers 879
Our sample period is from January 1995 to December 2010 We focus onthe post-1994 period because TASS started reporting information on ldquodefunctrdquofunds only after 19945 We exclude managed futuresCTAs and funds-of-fundswhich have different incentive fees and are likely to have different inflow-performance relations from typical individual hedge funds We also excludeclosed-end hedge funds as subscriptions in these funds are only possible duringthe initial issuing period so future flows are not possible This initial filterleaves us with 4939 open-end hedge funds6
We drop funds for which TASS does not contain information on organizationalcharacteristics such as management fees incentive fees and HWM provisionsIn addition we consider only funds with an uninterrupted series of net assetvalues and returns so that we can calculate inflows Further we restrict oursample to funds with at least 12 consecutive monthly returns available duringthe sample period If a fund stops reporting returns and then resumes at afuture date we use only the first sequence of uninterrupted data Finally weexclude funds with an incentive fee of zero because there can be no direct payfor performance for these funds
We conduct the analysis using quarterly data because we wish to include onlylagged (not contemporaneous) returns in the flow-performance specificationsand to have a relatively short gap between returns and flows7 To construct aquarterly sample we drop all fund-calendar quarters that have return infor-mation only for a fraction of the period We also require that a sample fundhave a subscription period less than or equal to three months so that quarterlyinflows are not restricted This sample construction process leaves us with asample of 2998 funds (50333 fund-quarter observations)
Table I presents descriptive statistics Time-varying variables such as Quar-terly flow and Quarterly returns are measured at the fund-quarter level andother contractual characteristics such as Management fees and Incentive feesare measured at the fund level8 All time-varying variables except fund age(Age) are winsorized at the 1st and 99th percentiles to minimize the effect ofoutliers
5 Defunct funds include funds that are liquidated merged or restructured as well as those thatstopped reporting returns to TASS (Fung et al (2008))
6 Some funds are closed to new investors but unfortunately we do not know whether a particularfund is taking new money at any point in time so we cannot exclude funds on the basis of thispolicy Including closed funds causes us to understate the flow-performance relation for funds thatare not closed
7 Prior versions of this paper presented estimates using annual data with contemporaneousannual returns included in the flow-performance specifications (Lim Sensoy and Weisbach (2013))The estimates of indirect incentives using annual data are very close to those reported here
8 TASS provides information on fundsrsquo organizational characteristics as of the last available dateof fund data Like most previous studies we also assume that these organizational characteristicsdo not change throughout the life of the fund Agarwal Daniel and Naik (2009) argue that fundsrsquoorganizational characteristics are unlikely to change much over time based on their discussionswith practitioners which suggests that if a manager wants to impose new contractual terms it iseasier for him to start a new fund with different terms than to go through the legal complicationsof changing an existing contract
880 The Journal of Finance Rcopy
Mean Quarterly flow is 80 with a median of 03 so the distributionis highly skewed Mean Quarterly returns is 26 the median is 21 Aver-age fund size (AUM) is $2108 million The remaining variables reflect time-invariant contractual features Summary statistics on these characteristics arevery close to those reported in prior studies (eg Agarwal Daniel and Naik(2009) Baquero and Verbeek (2009) Aragon and Nanda (2012)) The variableManagement fees is the annual percentage of fund assets (AUM) received bythe manager as compensation and has a sample mean (median) of 15 (15)The variable Incentive fees is the percentage of profits above the HWM receivedby the manager as compensation and has a sample mean (median) of 193(20) Over two-thirds of our sample funds 694 have a HWM provision685 report that they use leverage 191 are open to public investors andabout a quarter are on-shore funds
The fee data consist of the fees that are currently publicly quoted by thefunds These data could potentially misrepresent the true fees relevant forour calculations for three reasons First funds sometimes provide fee reduc-tions to particular strategic investors that are not reflected in the database(Ramadorai and Streatfield (2011)) Although we cannot investigate this pos-sibility directly it is unlikely to have a major impact on our conclusions Forexample if the true incentive fee (management fee) averaged over all investorswere 10 lower than what is reported in the database our estimates of indirectincentives using the base GIR model would be overstated by about 1 (5) Asecond issue is that fees can change over time Agarwal and Ray (2012) andDueskar et al (2012) find that fee changes are infrequent and tend to reflectpast performance when they do occur so that fee increases follow good perfor-mance and fee decreases follow poor performance This effect is magnified inindirect incentives because good performance today leads not only to inflowsbut also to higher proportional fees on those inflows Third it is possible thatthere are fee changes unobservable to researchers To the extent that observ-able and unobservable fee changes are positively correlated this possibilityleads us to understate indirect incentives
We consider three variables that reflect potential restrictions on the be-havior of flows First we use Total redemption period defined as the sum ofthe notice period and the redemption period where the notice period is thetime the investor has to give notice to the fund about an intention to with-draw money from the fund and the redemption period is the time that thefund takes to return the money after the notice period is over The Lockupperiod is the minimum time that an investor has to wait before withdrawinginvested money The Subscription period is the time delay between investingin a fund and actually purchasing fund shares The mean total redemption pe-riod lock-up period and subscription period are 109 097 and 036 quartersrespectively
Indirect Incentives of Hedge Fund Managers 881
III Estimating the Sensitivity of Fund Inflows to Performance
To understand the impact of performance on fund flows we employ aBayesian learning framework that presumes investors are continually eval-uating managers trying to assess their abilities (see Berk and Green (2004)Chung et al (2012)) A fundrsquos performance provides information about the man-agerrsquos ability so an observation of performance will lead investors to updatetheir assessment of his ability and allocate more capital to a fund if their es-timate of the managerrsquos ability increases The magnitude of the updates andhence the sensitivity of inflows to performance should depend on the infor-mativeness of the signal relative to the precision of the prior estimate of thefund managerrsquos ability The sensitivity of inflows to performance should alsodepend on the extent to which ability can be scaled to replicate a fundrsquos returndistribution on new capital
Measuring the indirect incentives of hedge fund managers requires an es-timate of the relation between fund performance and future inflows A longliterature beginning with Ippolito (1992) estimates this relation to be rela-tively strong in the mutual fund industry9 It is also positive in the privateequity industry (see eg Kaplan and Schoar (2005)) However the results forhedge funds are less clear Goetzmann Ingersoll and Ross (2003) report a neg-ative and concave relation whereas other studies including Agarwal Danieland Naik (2003) Fung et al (2008) Baquero and Verbeek (2009) and Dinget al (2009) find a positive one
A Empirical Specification
We estimate the following specification
Flowit = β0 +11sumj=1
β0+ j Returnitminus j + γ Xtminus1 + λY + Fixed effects + εit (1)
where Flowit represents flows for fund i in quarter t10
Following the literature on flows to mutual funds (eg Sirri and Tufano(1998) Chevalier and Ellison (1999)) or hedge funds (eg Fung et al (2008)Agarwal Daniel and Naik (2009)) we compute quarterly flows of capital intoa fund as follows
Flowit = AUMit minus AUMitminus1(1 + Returnit
)AUMitminus1
(2)
where AUMit and AUMitminus1 are the AUM of fund i at the end of quarters t andtminus1 respectively and Returnit is the net-of-fee return for fund i in quarter t
9 See Brown Halow and Starks (1996) Sirri and Tufano (1998) Chevalier and Ellison (1999)Barclay Pearson and Weisbach (1998) Del Guercio and Tkac (2002) Bollen (2007) Huang Weiand Yan (2007) and Sensoy (2009)
10 We restrict our estimates to quarterly specifications but prior versions of this paper alsoinclude annual and monthly specifications with similar results to those reported below
882 The Journal of Finance Rcopy
This definition expresses flows as a fraction of beginning-of-period (end ofprior period) AUM which is a natural benchmark from the perspective of afund manager assessing his or her incentives going forward For instance theoption deltas that comprise direct incentives are defined in terms of beginning-of-period AUM An alternative approach to computing fund flows is to scalethe denominator of equation (2) by (1 + Returnit) which expresses flows as afraction of what end-of-period AUM would have been in the absence of flowsAlthough this alternative definition is less intuitive for our purpose it leads tosimilar estimates of indirect incentives
The vector X in equation (1) consists of time-varying fund characteristicsthat include lagged flows the natural logarithm of AUM for fund i at tminus1 thenatural logarithm of one plus fund irsquos age in quarters at tminus1 and fund volatil-ity over the previous 12 months11 The vector Y includes time-invariant fundcharacteristics that include Management fees Incentive fees Total redemptionperiod Lockup period Subscription period and a set of indicator variables thatequal one if the fund has an HWM provision uses leverage is open to publicinvestors and is an on-shore fund All specifications include fixed effects forthe nine styles listed in Table I interacted with calendar quarter fixed effectsThese time-by-style effects capture all shocks observed or unobserved that arecommon to funds of a given style in a given quarter including the returns topeer funds and inflows to other funds of the same style12 Reported standarderrors are robust to heteroskedasticity and account for double clustering byfund and time period13
B Estimates of the Flow-Performance Relation
We present estimates of the flow-performance relation in Panel A ofTable II In Column (1) we include returns in the 11 quarters prior to thecurrent quarter In Column (2) we add a number of fund-level controls In eachspecification the coefficients on returns are positive and statistically signif-icant in most cases and decline sharply over time so the coefficient on themost recent quarterrsquos return is the largest If we sum the coefficients on the 11prior quarters the sum in Column (1) is 144 and in Column (2) is 150 Thesecoefficients imply that a one-standard-deviation increase in returns (9) willlead to about a 13 increase in fund size
11 For young funds that have for example only one yearrsquos worth of return history we cannotcompute lagged returns and flows In such cases we ldquodummy outrdquo missing lagged variables toretain observations To do so we set missing values of lagged flows and returns to zero and includean indicator for missing values
12 Time-by-style fixed effects perform the same adjustment as a factor model regression underthe assumption that the factor loadings are the same for all funds of a given style within a giventime period
13 We focus on linear specifications Although there is some evidence of nonlinear effects in thedata (using splines quadratics etc) they are small in magnitude and have little impact on theestimates obtained with linear specifications
Indirect Incentives of Hedge Fund Managers 883
Table IIEstimates of Flow-Performance Sensitivity
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) using quarterly data The dependent variable is Quarterly flow as defined byequation (2) See Table I for definitions of all independent variables The sample period is from 1995to 2010 We only employ the funds with a subscription period less than or equal to three months toavoid quarterly inflows being restricted by subscription periods Equation (1) is augmented in PanelB to include interaction terms of the log of the fundrsquos age plus one with prior performance and inPanel C to include interaction terms of Constrained with performance variables All specificationsin Panel A and B include style fixed effects quarter fixed effects and quarter-by-style fixed effectsOnly quarter fixed effects are included in Panel C due to multicollinearity Standard errors aredouble clustered by fund and year and indicate statistical significance at the 1 5and 10 level respectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel A Base-case
Quarterly returns(tminus1) 0501 (0045) 0496 (0043)Quarterly returns(tminus2) 0314 (0039) 0263 (0036)Quarterly returns(tminus3) 0209 (0034) 0190 (0028)Quarterly returns(tminus4) 0117 (0035) 0111 (0030)Quarterly returns(tminus5) 0089 (0030) 0098 (0027)Quarterly returns(tminus6) 0054 (0033) 0066 (0030)Quarterly returns(tminus7) 0083 (0027) 0103 (0025)Quarterly returns(tminus8) 0044 (0022) 0060 (0022)Quarterly returns(tminus9) 0012 (0029) 0034 (0029)Quarterly returns(tminus10) minus0010 (0030) 0019 (0029)Quarterly returns(tminus11) 0025 (0027) 0056 (0026)Quarterly flow(tminus1) 0158 (0008)Log(Age + 1) minus0004 (0005)Log(AUM(tminus1)+1) minus0024 (0001)Volatility minus0141 (0037)Management fees 0211 (0386)Incentive fees 0005 (0053)HWM indicator 0020 (0004)Leveraged indicator 0000 (0004)Open-to-public indicator minus0002 (0005)On-shore indicator minus0032 (0005)Log(Total redemption period +1) 0034 (0007)Log(Lock-up period +1) minus0005 (0003)Log(Subscription period +1) minus0023 (0015)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0162 0194F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1940 (0000) 2077 (0000)
(Continued)
884 The Journal of Finance Rcopy
Table IImdashContinued
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel B Age interactions
Quarterly returns(tminus1)timesLog(Age in quarters+1) minus0319 (0035) minus0285 (0034)Quarterly returns(tminus2)timesLog(Age in quarters+1) minus0171 (0027) minus0125 (0025)Quarterly returns(tminus3)timesLog(Age in quarters+1) minus0051 (0036) minus0033 (0036)Quarterly returns(tminus4)timesLog(Age in quarters+1) minus0003 (0032) 0001 (0032)Quarterly returns(tminus5)timesLog(Age in quarters+1) minus0041 (0030) minus0042 (0029)Quarterly returns(tminus6)timesLog(Age in quarters+1) 0038 (0033) 0040 (0031)Quarterly returns(tminus7)timesLog(Age in quarters+1) 0013 (0033) 0007 (0032)Quarterly returns(tminus8)timesLog(Age in quarters+1) 0033 (0033) 0034 (0033)Quarterly returns(tminus9)timesLog(Age in quarters+1) 0003 (0040) minus0004 (0040)Quarterly returns(tminus10)timesLog(Age in quarters+1) 0132 (0038) 0133 (0035)Quarterly returns(tminus11)timesLog(Age in quarters+1) 0018 (0040) 0007 (0041)Quarterly returns(tminus1) 1325 (0117) 1233 (0113)Quarterly returns(tminus2) 0773 (0095) 0602 (0087)Quarterly returns(tminus3) 0354 (0114) 0287 (0110)Quarterly returns(tminus4) 0124 (0112) 0110 (0106)Quarterly returns(tminus5) 0200 (0100) 0215 (0092)Quarterly returns(tminus6) minus0067 (0112) minus0056 (0104)Quarterly returns(tminus7) 0033 (0106) 0079 (0099)Quarterly returns(tminus8) minus0066 (0107) minus0046 (0107)Quarterly returns(tminus9) 0004 (0128) 0053 (0127)Quarterly returns(tminus10) minus0416 (0122) minus0383 (0114)Quarterly returns(tminus11) minus0038 (0126) 0037 (0130)Log(Age + 1) minus0026 (0007) minus0000 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0168 0198F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1788 (0000) 1570 (0000)All age interaction terms 1214 (0000) 1084 (0000)
Panel C Scalability interactions
Quarterly returns(tminus1)timesConstrained minus0215 (0048) minus0215 (0047)Quarterly returns(tminus2)timesConstrained minus0096 (0043) minus0078 (0039)Quarterly returns(tminus3)timesConstrained minus0102 (0038) minus0107 (0037)Quarterly returns(tminus4)timesConstrained 0013 (0034) 0013 (0036)Quarterly returns(tminus5)timesConstrained minus0009 (0032) minus0022 (0032)Quarterly returns(tminus6)timesConstrained 0026 (0034) 0013 (0034)Quarterly returns(tminus7)timesConstrained minus0044 (0036) minus0057 (0036)Quarterly returns(tminus8)timesConstrained 0032 (0037) 0033 (0036)Quarterly returns(tminus9)timesConstrained minus0085 (0030) minus0090 (0031)Quarterly returns(tminus10)timesConstrained minus0023 (0036) minus0007 (0036)Quarterly returns(tminus11)timesConstrained minus0007 (0037) minus0004 (0038)
(Continued)
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 875
relation Section IV estimates managersrsquo direct and indirect incentives SectionV discusses implications of our estimates Section VI concludes
I Quantifying the Magnitude of Pay for Performance of HedgeFund Managers
A Direct Pay for Performance
Hedge fund managersrsquo compensation generally consists of management feeswhich are a percentage of AUM (often around 15) plus incentive fees whichare a percentage (usually 20) of profits or of profits earned above the HWMIn addition hedge fund managers usually make a personal investment in thefund The direct pay for performance that a manager receives comes from theincentive fees as well as his personal investment in the fund both of whichincrease in value with the fundrsquos performance Quantifying these direct perfor-mance incentives is complicated because of the option-like features containedin the hedge fund managerrsquos incentive fee contract In particular the incen-tive fee contract resembles a portfolio of call options one per investor in thefund The exercise price of each option is determined by the investorrsquos time ofentrance into the fund the fundrsquos hurdle rate and the historical HWM levelpertaining to the investorrsquos assets Thus even if different managers have thesame 20 incentive fee rate their actual direct pay-performance sensitivitywill vary depending on the distance between the current asset value and theexercise prices of these options
To estimate the direct pay-performance sensitivity we use Agarwal Danieland Noikrsquos (2009) total delta approach which measures the impact of an in-cremental one-percentage-point return to fund investors on the value of themanagerrsquos incentive fees plus the increase in the value of the managerrsquos ownownership stake The total delta of the managerrsquos claim to the fund is equalto the sum of these individual optionsrsquo deltas plus the delta of the managerrsquospersonal stake in the fund We follow Agarwal Daniel and Naikrsquos (2009) andassume that the managerrsquos initial stake is zero but the stake grows over timeas managers reinvest all of their incentive fees back into the fund1 Details ofthis calculation are described in Appendix A
B Indirect Pay for Performance
In addition to the pay for performance from incentive fees and their owninvestment in the fund hedge fund managersrsquo lifetime income changes withperformance through a reputational effect good performance increases themarketrsquos perception of a managerrsquos ability leading to higher inflows of newinvestment to the fund Ultimately a fundrsquos managers will receive part of
1 Assuming instead that the managerrsquos initial stake is 1 or 2 of AUM increases the estimatesof indirect incentives by 5 and 10 respectively and the estimates of direct incentives by 2and 5 respectively These changes do not affect our conclusion that indirect incentives are largerelative to direct incentives
876 The Journal of Finance Rcopy
these inflows as future management and incentive fees Furthermore goodperformance mechanically increases the value of existing investorsrsquo stakes inthe fund A portion of this increase will likewise be paid over time in future feesto the fund manager The expectation of this future income depends on todayrsquosperformance leading to what we refer to as indirect incentives
To evaluate the magnitude of these indirect incentives we first need esti-mates of the way in which performance affects expected inflows to the fundThese estimates are discussed in Section III below We also need a model ofthe present value of a managerrsquos expected lifetime compensation as a fractionof fund assets This model should predict for each incremental dollar undermanagement the increase in the managerrsquos expected compensation over thefuture lifetime of the fund We use four such models
The first model is that of GIR The GIR model leads to the lowest estimatesof the sensitivity of future income to todayrsquos performance so it can be thoughtof as providing a lower bound on our estimates of indirect incentives The GIRmodel is a contingent-claims model of the fraction of fund assets that accrue tothe fund manager in expected future compensation Key features of this modelare that compensation is a fixed management fee plus a percentage of profitsabove a HWM the fundrsquos asset value follows a martingale with drift generatedby the managerrsquos alpha investors continuously withdraw assets (eg an en-dowment investor might withdraw 5 per year) and an investor liquidates hisor her position following a sufficiently negative return shock (so that expectedfund life is finite)
The GIR model does not account for all factors that could affect managersrsquofuture compensation and in turn indirect incentives In particular the GIRmodel does not allow the fundrsquos asset value to grow through future performance-based fund flows Under appropriate assumptions the effect of such flows inthe context of the GIR model is to increase the variance of the fundrsquos AUM2
For this reason our second model reports estimates from a variant of the GIRmodel in which fund variance is augmented
Another factor not accounted for by the GIR model is the fact that a fundrsquosportfolio is endogenously chosen with managers adjusting their portfolios tomaximize their incomes given a particular incentive scheme Following LWYour third model allows managers to perform such adjustments by levering theirfunds strategically Given that hedge fund managers typically do use leverageestimates of indirect incentives that incorporate this feature are likely to bemore accurate
Finally we present estimates from a fourth model also introduced by LWYthat augments the basic LWY model to include future performance-based fundflows Note also that the LWY model nests the GIR model as a special case in
2 Suppose as an approximation that the flow-performance relationship is linear in logs andflows are contemporaneous with returns Suppose that without flows the log AUM of the fundevolves as a martingale as in the GIR model st+1 = st + et+1 With performance-based flows st+1 =st + get+1 where g gt 1 captures the flow effect Therefore even with performance-based flows thelog AUM would still follow a martingale so the GIR model still applies but with a higher variancethan in the case of no flows We thank the Associate Editor for suggesting this argument to us
Indirect Incentives of Hedge Fund Managers 877
which leverage is zero so all four sets of estimates can be viewed as comingfrom variations of the LWY model
Using different models to estimate indirect incentives allows us to gaugethe importance of different factors especially the importance of futureperformance-based fund flows and endogenous portfolio choice in determin-ing the fraction of a fundrsquos value that will ultimately accrue to managers infuture fees3 Details on all four models our choices of model parameters andthe associated calculations are described in Appendix B
For each of the models above we estimate the present value of the total(management plus incentive) future fees that the manager earns on an extradollar of AUM To calculate indirect pay for performance we multiply thispresent value by an estimate of the number of incremental dollars of AUMthat result from a one-percentage-point increase in returns to investors Thelatter increase consists of two parts namely the mechanical increase in thevalue of existing investorsrsquo stakes plus incremental inflows of new investmentIn this way the models for a managerrsquos fee value together with our estimatesof the flow-performance relation facing hedge fund managers provide an esti-mate of the present value of the incremental future revenue that the hedgefund manager expects to earn as a result of a one-percentage-point improve-ment in current net returns4 This calculation results in an incentive measuredenominated in dollars per percentage point change in returns We also con-duct the same calculation for a one-dollar improvement in investor return toobtain a measure in terms of dollars in managerial wealth per dollar returnedto shareholders Both of these measures are commonly used in the literatureon manager incentives
Because the GIR and LWY models estimate present values no further ad-justment for the riskiness of future income is required Also the estimates donot require that the manager continue to manage the fund in the future underthe assumption that the present value of the managerrsquos claims to future feeincome can be monetized when the manager departs
II Hedge Fund Data
Our data come from the TASS database which covers about 40 of thehedge fund universe (Agarwal Daniel and Naik (2009)) Summary statisticsfor key sample fund characteristics reported in Table I are very close to thosefor the sample considered by Agarwal Daniel and Naik (2009) who mergeand consolidate four major databases (CISDM HFR MSCI and TASS) Wetherefore conclude that our sample of hedge funds is representative of thehedge fund universe
3 In addition to GIR and LWY there have been a number of other attempts to value managersrsquoclaims to hedge funds Important contributions include Panageas and Westerfield (2009) Drechsler(2014) and Guasoni and Oblόj (2013)
4 Net returns can be improved either through increased gross returns or decreased costs borneby the fund such as financing costs security lending fees and settlement charges The incentiveswe measure are therefore incentives to achieve both
878 The Journal of Finance Rcopy
Table ISummary Statistics
This table presents descriptive statistics for the sample of 2998 funds during the 1995 to 2010sample period Time-varying variables are reported at the fund-quarter level and other time-invariant contractual characteristics are reported at the fund level Quarterly flow is the differencebetween the reported quarter-end AUM and the quarter-beginning AUM times (1 + Quarterlyreturns) scaled by quarter-beginning AUM Quarterly returns are the reported quarterly net-of-fee returns AUM is assets under management in millions measured at the end of each quarterAge is the number of quarters from the fundrsquos inception date measured at the beginning of eachquarter Volatility is the standard deviation of monthly (net-of-fee) returns over the 12 monthsprior to each quarter multiplied by the square root of three to arrive at a quarterly figure Alltime-varying variables except Age are winsorized at the 1st and 99th percentiles Management feesare the percentage of the assets charged by the fund as regular fees in annual terms Incentive feesare the percentage of positive profits received by the manager as performance incentives HWMis an indicator variable that equals one if the fund has a HWM provision and zero otherwiseLeverage is an indicator variable that equals one if the fund uses leverage and zero otherwiseOpen-to-public is an indicator variable that equals one if the fund allows public investment andzero otherwise On-shore is an indicator variable that equals one if the fund is domiciled in theUnited States and zero otherwise Total redemption period is the sum of the notice period and theredemption period measured in quarters where the notice period is the time the investor has togive notice to the fund about an intention to withdraw money from the fund and the redemptionperiod is the time that the fund takes to return the money after the notice period is over Lockupperiod is the minimum time in quarters that an investor has to wait before withdrawing investedmoney Lockup period is considered to be zero for a fund that has no lock-up provision Subscriptionperiod is the time delay measured in quarters between investing in a fund and actually purchasingfund shares Constrained is an indicator variable that equals one if the fundrsquos strategy belongs toConvertible Arbitrage Fixed Income Arbitrage Emerging Markets and Event Driven strategiesand zero otherwise (Ding et al (2009))
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 50333 798 minus413 034 1014 3701Quarterly returns () 50333 256 minus091 211 566 905AUM ($ million) 50333 2108 110 393 1300 6370Age (quarters) 50333 1756 609 1319 2434 1532Volatility () 50333 533 300 441 804 243Management fees ( annualized) 2998 15 10 15 20 05Incentive fees () 2998 193 200 200 200 34HWM 2998 0694 0000 1000 1000 0461Leverage 2998 0685 0000 1000 1000 0464Open-to-public 2998 0191 0000 0000 0000 0394On-shore 2998 0230 0000 0000 0000 0421Total redemption period (quarters) 2998 1085 0556 0667 1333 1015Lockup period (quarters) 2998 0974 0000 0000 0000 2077Subscription period (quarters) 2998 0356 0333 0333 0333 0190Style
Constrained 2998 0303 0000 0000 1000 0460Constrained = 1
Convertible Arbitrage 2998 0034Emerging Markets 2998 0120Event Driven 2998 0096Fixed Income Arbitrage 2998 0052
Constrained = 0Equity Market Neutral 2998 0073Global Macro 2998 0078LongShort Equity Hedge 2998 0420Multi-Strategy 2998 0077Other 2998 0049
Indirect Incentives of Hedge Fund Managers 879
Our sample period is from January 1995 to December 2010 We focus onthe post-1994 period because TASS started reporting information on ldquodefunctrdquofunds only after 19945 We exclude managed futuresCTAs and funds-of-fundswhich have different incentive fees and are likely to have different inflow-performance relations from typical individual hedge funds We also excludeclosed-end hedge funds as subscriptions in these funds are only possible duringthe initial issuing period so future flows are not possible This initial filterleaves us with 4939 open-end hedge funds6
We drop funds for which TASS does not contain information on organizationalcharacteristics such as management fees incentive fees and HWM provisionsIn addition we consider only funds with an uninterrupted series of net assetvalues and returns so that we can calculate inflows Further we restrict oursample to funds with at least 12 consecutive monthly returns available duringthe sample period If a fund stops reporting returns and then resumes at afuture date we use only the first sequence of uninterrupted data Finally weexclude funds with an incentive fee of zero because there can be no direct payfor performance for these funds
We conduct the analysis using quarterly data because we wish to include onlylagged (not contemporaneous) returns in the flow-performance specificationsand to have a relatively short gap between returns and flows7 To construct aquarterly sample we drop all fund-calendar quarters that have return infor-mation only for a fraction of the period We also require that a sample fundhave a subscription period less than or equal to three months so that quarterlyinflows are not restricted This sample construction process leaves us with asample of 2998 funds (50333 fund-quarter observations)
Table I presents descriptive statistics Time-varying variables such as Quar-terly flow and Quarterly returns are measured at the fund-quarter level andother contractual characteristics such as Management fees and Incentive feesare measured at the fund level8 All time-varying variables except fund age(Age) are winsorized at the 1st and 99th percentiles to minimize the effect ofoutliers
5 Defunct funds include funds that are liquidated merged or restructured as well as those thatstopped reporting returns to TASS (Fung et al (2008))
6 Some funds are closed to new investors but unfortunately we do not know whether a particularfund is taking new money at any point in time so we cannot exclude funds on the basis of thispolicy Including closed funds causes us to understate the flow-performance relation for funds thatare not closed
7 Prior versions of this paper presented estimates using annual data with contemporaneousannual returns included in the flow-performance specifications (Lim Sensoy and Weisbach (2013))The estimates of indirect incentives using annual data are very close to those reported here
8 TASS provides information on fundsrsquo organizational characteristics as of the last available dateof fund data Like most previous studies we also assume that these organizational characteristicsdo not change throughout the life of the fund Agarwal Daniel and Naik (2009) argue that fundsrsquoorganizational characteristics are unlikely to change much over time based on their discussionswith practitioners which suggests that if a manager wants to impose new contractual terms it iseasier for him to start a new fund with different terms than to go through the legal complicationsof changing an existing contract
880 The Journal of Finance Rcopy
Mean Quarterly flow is 80 with a median of 03 so the distributionis highly skewed Mean Quarterly returns is 26 the median is 21 Aver-age fund size (AUM) is $2108 million The remaining variables reflect time-invariant contractual features Summary statistics on these characteristics arevery close to those reported in prior studies (eg Agarwal Daniel and Naik(2009) Baquero and Verbeek (2009) Aragon and Nanda (2012)) The variableManagement fees is the annual percentage of fund assets (AUM) received bythe manager as compensation and has a sample mean (median) of 15 (15)The variable Incentive fees is the percentage of profits above the HWM receivedby the manager as compensation and has a sample mean (median) of 193(20) Over two-thirds of our sample funds 694 have a HWM provision685 report that they use leverage 191 are open to public investors andabout a quarter are on-shore funds
The fee data consist of the fees that are currently publicly quoted by thefunds These data could potentially misrepresent the true fees relevant forour calculations for three reasons First funds sometimes provide fee reduc-tions to particular strategic investors that are not reflected in the database(Ramadorai and Streatfield (2011)) Although we cannot investigate this pos-sibility directly it is unlikely to have a major impact on our conclusions Forexample if the true incentive fee (management fee) averaged over all investorswere 10 lower than what is reported in the database our estimates of indirectincentives using the base GIR model would be overstated by about 1 (5) Asecond issue is that fees can change over time Agarwal and Ray (2012) andDueskar et al (2012) find that fee changes are infrequent and tend to reflectpast performance when they do occur so that fee increases follow good perfor-mance and fee decreases follow poor performance This effect is magnified inindirect incentives because good performance today leads not only to inflowsbut also to higher proportional fees on those inflows Third it is possible thatthere are fee changes unobservable to researchers To the extent that observ-able and unobservable fee changes are positively correlated this possibilityleads us to understate indirect incentives
We consider three variables that reflect potential restrictions on the be-havior of flows First we use Total redemption period defined as the sum ofthe notice period and the redemption period where the notice period is thetime the investor has to give notice to the fund about an intention to with-draw money from the fund and the redemption period is the time that thefund takes to return the money after the notice period is over The Lockupperiod is the minimum time that an investor has to wait before withdrawinginvested money The Subscription period is the time delay between investingin a fund and actually purchasing fund shares The mean total redemption pe-riod lock-up period and subscription period are 109 097 and 036 quartersrespectively
Indirect Incentives of Hedge Fund Managers 881
III Estimating the Sensitivity of Fund Inflows to Performance
To understand the impact of performance on fund flows we employ aBayesian learning framework that presumes investors are continually eval-uating managers trying to assess their abilities (see Berk and Green (2004)Chung et al (2012)) A fundrsquos performance provides information about the man-agerrsquos ability so an observation of performance will lead investors to updatetheir assessment of his ability and allocate more capital to a fund if their es-timate of the managerrsquos ability increases The magnitude of the updates andhence the sensitivity of inflows to performance should depend on the infor-mativeness of the signal relative to the precision of the prior estimate of thefund managerrsquos ability The sensitivity of inflows to performance should alsodepend on the extent to which ability can be scaled to replicate a fundrsquos returndistribution on new capital
Measuring the indirect incentives of hedge fund managers requires an es-timate of the relation between fund performance and future inflows A longliterature beginning with Ippolito (1992) estimates this relation to be rela-tively strong in the mutual fund industry9 It is also positive in the privateequity industry (see eg Kaplan and Schoar (2005)) However the results forhedge funds are less clear Goetzmann Ingersoll and Ross (2003) report a neg-ative and concave relation whereas other studies including Agarwal Danieland Naik (2003) Fung et al (2008) Baquero and Verbeek (2009) and Dinget al (2009) find a positive one
A Empirical Specification
We estimate the following specification
Flowit = β0 +11sumj=1
β0+ j Returnitminus j + γ Xtminus1 + λY + Fixed effects + εit (1)
where Flowit represents flows for fund i in quarter t10
Following the literature on flows to mutual funds (eg Sirri and Tufano(1998) Chevalier and Ellison (1999)) or hedge funds (eg Fung et al (2008)Agarwal Daniel and Naik (2009)) we compute quarterly flows of capital intoa fund as follows
Flowit = AUMit minus AUMitminus1(1 + Returnit
)AUMitminus1
(2)
where AUMit and AUMitminus1 are the AUM of fund i at the end of quarters t andtminus1 respectively and Returnit is the net-of-fee return for fund i in quarter t
9 See Brown Halow and Starks (1996) Sirri and Tufano (1998) Chevalier and Ellison (1999)Barclay Pearson and Weisbach (1998) Del Guercio and Tkac (2002) Bollen (2007) Huang Weiand Yan (2007) and Sensoy (2009)
10 We restrict our estimates to quarterly specifications but prior versions of this paper alsoinclude annual and monthly specifications with similar results to those reported below
882 The Journal of Finance Rcopy
This definition expresses flows as a fraction of beginning-of-period (end ofprior period) AUM which is a natural benchmark from the perspective of afund manager assessing his or her incentives going forward For instance theoption deltas that comprise direct incentives are defined in terms of beginning-of-period AUM An alternative approach to computing fund flows is to scalethe denominator of equation (2) by (1 + Returnit) which expresses flows as afraction of what end-of-period AUM would have been in the absence of flowsAlthough this alternative definition is less intuitive for our purpose it leads tosimilar estimates of indirect incentives
The vector X in equation (1) consists of time-varying fund characteristicsthat include lagged flows the natural logarithm of AUM for fund i at tminus1 thenatural logarithm of one plus fund irsquos age in quarters at tminus1 and fund volatil-ity over the previous 12 months11 The vector Y includes time-invariant fundcharacteristics that include Management fees Incentive fees Total redemptionperiod Lockup period Subscription period and a set of indicator variables thatequal one if the fund has an HWM provision uses leverage is open to publicinvestors and is an on-shore fund All specifications include fixed effects forthe nine styles listed in Table I interacted with calendar quarter fixed effectsThese time-by-style effects capture all shocks observed or unobserved that arecommon to funds of a given style in a given quarter including the returns topeer funds and inflows to other funds of the same style12 Reported standarderrors are robust to heteroskedasticity and account for double clustering byfund and time period13
B Estimates of the Flow-Performance Relation
We present estimates of the flow-performance relation in Panel A ofTable II In Column (1) we include returns in the 11 quarters prior to thecurrent quarter In Column (2) we add a number of fund-level controls In eachspecification the coefficients on returns are positive and statistically signif-icant in most cases and decline sharply over time so the coefficient on themost recent quarterrsquos return is the largest If we sum the coefficients on the 11prior quarters the sum in Column (1) is 144 and in Column (2) is 150 Thesecoefficients imply that a one-standard-deviation increase in returns (9) willlead to about a 13 increase in fund size
11 For young funds that have for example only one yearrsquos worth of return history we cannotcompute lagged returns and flows In such cases we ldquodummy outrdquo missing lagged variables toretain observations To do so we set missing values of lagged flows and returns to zero and includean indicator for missing values
12 Time-by-style fixed effects perform the same adjustment as a factor model regression underthe assumption that the factor loadings are the same for all funds of a given style within a giventime period
13 We focus on linear specifications Although there is some evidence of nonlinear effects in thedata (using splines quadratics etc) they are small in magnitude and have little impact on theestimates obtained with linear specifications
Indirect Incentives of Hedge Fund Managers 883
Table IIEstimates of Flow-Performance Sensitivity
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) using quarterly data The dependent variable is Quarterly flow as defined byequation (2) See Table I for definitions of all independent variables The sample period is from 1995to 2010 We only employ the funds with a subscription period less than or equal to three months toavoid quarterly inflows being restricted by subscription periods Equation (1) is augmented in PanelB to include interaction terms of the log of the fundrsquos age plus one with prior performance and inPanel C to include interaction terms of Constrained with performance variables All specificationsin Panel A and B include style fixed effects quarter fixed effects and quarter-by-style fixed effectsOnly quarter fixed effects are included in Panel C due to multicollinearity Standard errors aredouble clustered by fund and year and indicate statistical significance at the 1 5and 10 level respectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel A Base-case
Quarterly returns(tminus1) 0501 (0045) 0496 (0043)Quarterly returns(tminus2) 0314 (0039) 0263 (0036)Quarterly returns(tminus3) 0209 (0034) 0190 (0028)Quarterly returns(tminus4) 0117 (0035) 0111 (0030)Quarterly returns(tminus5) 0089 (0030) 0098 (0027)Quarterly returns(tminus6) 0054 (0033) 0066 (0030)Quarterly returns(tminus7) 0083 (0027) 0103 (0025)Quarterly returns(tminus8) 0044 (0022) 0060 (0022)Quarterly returns(tminus9) 0012 (0029) 0034 (0029)Quarterly returns(tminus10) minus0010 (0030) 0019 (0029)Quarterly returns(tminus11) 0025 (0027) 0056 (0026)Quarterly flow(tminus1) 0158 (0008)Log(Age + 1) minus0004 (0005)Log(AUM(tminus1)+1) minus0024 (0001)Volatility minus0141 (0037)Management fees 0211 (0386)Incentive fees 0005 (0053)HWM indicator 0020 (0004)Leveraged indicator 0000 (0004)Open-to-public indicator minus0002 (0005)On-shore indicator minus0032 (0005)Log(Total redemption period +1) 0034 (0007)Log(Lock-up period +1) minus0005 (0003)Log(Subscription period +1) minus0023 (0015)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0162 0194F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1940 (0000) 2077 (0000)
(Continued)
884 The Journal of Finance Rcopy
Table IImdashContinued
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel B Age interactions
Quarterly returns(tminus1)timesLog(Age in quarters+1) minus0319 (0035) minus0285 (0034)Quarterly returns(tminus2)timesLog(Age in quarters+1) minus0171 (0027) minus0125 (0025)Quarterly returns(tminus3)timesLog(Age in quarters+1) minus0051 (0036) minus0033 (0036)Quarterly returns(tminus4)timesLog(Age in quarters+1) minus0003 (0032) 0001 (0032)Quarterly returns(tminus5)timesLog(Age in quarters+1) minus0041 (0030) minus0042 (0029)Quarterly returns(tminus6)timesLog(Age in quarters+1) 0038 (0033) 0040 (0031)Quarterly returns(tminus7)timesLog(Age in quarters+1) 0013 (0033) 0007 (0032)Quarterly returns(tminus8)timesLog(Age in quarters+1) 0033 (0033) 0034 (0033)Quarterly returns(tminus9)timesLog(Age in quarters+1) 0003 (0040) minus0004 (0040)Quarterly returns(tminus10)timesLog(Age in quarters+1) 0132 (0038) 0133 (0035)Quarterly returns(tminus11)timesLog(Age in quarters+1) 0018 (0040) 0007 (0041)Quarterly returns(tminus1) 1325 (0117) 1233 (0113)Quarterly returns(tminus2) 0773 (0095) 0602 (0087)Quarterly returns(tminus3) 0354 (0114) 0287 (0110)Quarterly returns(tminus4) 0124 (0112) 0110 (0106)Quarterly returns(tminus5) 0200 (0100) 0215 (0092)Quarterly returns(tminus6) minus0067 (0112) minus0056 (0104)Quarterly returns(tminus7) 0033 (0106) 0079 (0099)Quarterly returns(tminus8) minus0066 (0107) minus0046 (0107)Quarterly returns(tminus9) 0004 (0128) 0053 (0127)Quarterly returns(tminus10) minus0416 (0122) minus0383 (0114)Quarterly returns(tminus11) minus0038 (0126) 0037 (0130)Log(Age + 1) minus0026 (0007) minus0000 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0168 0198F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1788 (0000) 1570 (0000)All age interaction terms 1214 (0000) 1084 (0000)
Panel C Scalability interactions
Quarterly returns(tminus1)timesConstrained minus0215 (0048) minus0215 (0047)Quarterly returns(tminus2)timesConstrained minus0096 (0043) minus0078 (0039)Quarterly returns(tminus3)timesConstrained minus0102 (0038) minus0107 (0037)Quarterly returns(tminus4)timesConstrained 0013 (0034) 0013 (0036)Quarterly returns(tminus5)timesConstrained minus0009 (0032) minus0022 (0032)Quarterly returns(tminus6)timesConstrained 0026 (0034) 0013 (0034)Quarterly returns(tminus7)timesConstrained minus0044 (0036) minus0057 (0036)Quarterly returns(tminus8)timesConstrained 0032 (0037) 0033 (0036)Quarterly returns(tminus9)timesConstrained minus0085 (0030) minus0090 (0031)Quarterly returns(tminus10)timesConstrained minus0023 (0036) minus0007 (0036)Quarterly returns(tminus11)timesConstrained minus0007 (0037) minus0004 (0038)
(Continued)
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
876 The Journal of Finance Rcopy
these inflows as future management and incentive fees Furthermore goodperformance mechanically increases the value of existing investorsrsquo stakes inthe fund A portion of this increase will likewise be paid over time in future feesto the fund manager The expectation of this future income depends on todayrsquosperformance leading to what we refer to as indirect incentives
To evaluate the magnitude of these indirect incentives we first need esti-mates of the way in which performance affects expected inflows to the fundThese estimates are discussed in Section III below We also need a model ofthe present value of a managerrsquos expected lifetime compensation as a fractionof fund assets This model should predict for each incremental dollar undermanagement the increase in the managerrsquos expected compensation over thefuture lifetime of the fund We use four such models
The first model is that of GIR The GIR model leads to the lowest estimatesof the sensitivity of future income to todayrsquos performance so it can be thoughtof as providing a lower bound on our estimates of indirect incentives The GIRmodel is a contingent-claims model of the fraction of fund assets that accrue tothe fund manager in expected future compensation Key features of this modelare that compensation is a fixed management fee plus a percentage of profitsabove a HWM the fundrsquos asset value follows a martingale with drift generatedby the managerrsquos alpha investors continuously withdraw assets (eg an en-dowment investor might withdraw 5 per year) and an investor liquidates hisor her position following a sufficiently negative return shock (so that expectedfund life is finite)
The GIR model does not account for all factors that could affect managersrsquofuture compensation and in turn indirect incentives In particular the GIRmodel does not allow the fundrsquos asset value to grow through future performance-based fund flows Under appropriate assumptions the effect of such flows inthe context of the GIR model is to increase the variance of the fundrsquos AUM2
For this reason our second model reports estimates from a variant of the GIRmodel in which fund variance is augmented
Another factor not accounted for by the GIR model is the fact that a fundrsquosportfolio is endogenously chosen with managers adjusting their portfolios tomaximize their incomes given a particular incentive scheme Following LWYour third model allows managers to perform such adjustments by levering theirfunds strategically Given that hedge fund managers typically do use leverageestimates of indirect incentives that incorporate this feature are likely to bemore accurate
Finally we present estimates from a fourth model also introduced by LWYthat augments the basic LWY model to include future performance-based fundflows Note also that the LWY model nests the GIR model as a special case in
2 Suppose as an approximation that the flow-performance relationship is linear in logs andflows are contemporaneous with returns Suppose that without flows the log AUM of the fundevolves as a martingale as in the GIR model st+1 = st + et+1 With performance-based flows st+1 =st + get+1 where g gt 1 captures the flow effect Therefore even with performance-based flows thelog AUM would still follow a martingale so the GIR model still applies but with a higher variancethan in the case of no flows We thank the Associate Editor for suggesting this argument to us
Indirect Incentives of Hedge Fund Managers 877
which leverage is zero so all four sets of estimates can be viewed as comingfrom variations of the LWY model
Using different models to estimate indirect incentives allows us to gaugethe importance of different factors especially the importance of futureperformance-based fund flows and endogenous portfolio choice in determin-ing the fraction of a fundrsquos value that will ultimately accrue to managers infuture fees3 Details on all four models our choices of model parameters andthe associated calculations are described in Appendix B
For each of the models above we estimate the present value of the total(management plus incentive) future fees that the manager earns on an extradollar of AUM To calculate indirect pay for performance we multiply thispresent value by an estimate of the number of incremental dollars of AUMthat result from a one-percentage-point increase in returns to investors Thelatter increase consists of two parts namely the mechanical increase in thevalue of existing investorsrsquo stakes plus incremental inflows of new investmentIn this way the models for a managerrsquos fee value together with our estimatesof the flow-performance relation facing hedge fund managers provide an esti-mate of the present value of the incremental future revenue that the hedgefund manager expects to earn as a result of a one-percentage-point improve-ment in current net returns4 This calculation results in an incentive measuredenominated in dollars per percentage point change in returns We also con-duct the same calculation for a one-dollar improvement in investor return toobtain a measure in terms of dollars in managerial wealth per dollar returnedto shareholders Both of these measures are commonly used in the literatureon manager incentives
Because the GIR and LWY models estimate present values no further ad-justment for the riskiness of future income is required Also the estimates donot require that the manager continue to manage the fund in the future underthe assumption that the present value of the managerrsquos claims to future feeincome can be monetized when the manager departs
II Hedge Fund Data
Our data come from the TASS database which covers about 40 of thehedge fund universe (Agarwal Daniel and Naik (2009)) Summary statisticsfor key sample fund characteristics reported in Table I are very close to thosefor the sample considered by Agarwal Daniel and Naik (2009) who mergeand consolidate four major databases (CISDM HFR MSCI and TASS) Wetherefore conclude that our sample of hedge funds is representative of thehedge fund universe
3 In addition to GIR and LWY there have been a number of other attempts to value managersrsquoclaims to hedge funds Important contributions include Panageas and Westerfield (2009) Drechsler(2014) and Guasoni and Oblόj (2013)
4 Net returns can be improved either through increased gross returns or decreased costs borneby the fund such as financing costs security lending fees and settlement charges The incentiveswe measure are therefore incentives to achieve both
878 The Journal of Finance Rcopy
Table ISummary Statistics
This table presents descriptive statistics for the sample of 2998 funds during the 1995 to 2010sample period Time-varying variables are reported at the fund-quarter level and other time-invariant contractual characteristics are reported at the fund level Quarterly flow is the differencebetween the reported quarter-end AUM and the quarter-beginning AUM times (1 + Quarterlyreturns) scaled by quarter-beginning AUM Quarterly returns are the reported quarterly net-of-fee returns AUM is assets under management in millions measured at the end of each quarterAge is the number of quarters from the fundrsquos inception date measured at the beginning of eachquarter Volatility is the standard deviation of monthly (net-of-fee) returns over the 12 monthsprior to each quarter multiplied by the square root of three to arrive at a quarterly figure Alltime-varying variables except Age are winsorized at the 1st and 99th percentiles Management feesare the percentage of the assets charged by the fund as regular fees in annual terms Incentive feesare the percentage of positive profits received by the manager as performance incentives HWMis an indicator variable that equals one if the fund has a HWM provision and zero otherwiseLeverage is an indicator variable that equals one if the fund uses leverage and zero otherwiseOpen-to-public is an indicator variable that equals one if the fund allows public investment andzero otherwise On-shore is an indicator variable that equals one if the fund is domiciled in theUnited States and zero otherwise Total redemption period is the sum of the notice period and theredemption period measured in quarters where the notice period is the time the investor has togive notice to the fund about an intention to withdraw money from the fund and the redemptionperiod is the time that the fund takes to return the money after the notice period is over Lockupperiod is the minimum time in quarters that an investor has to wait before withdrawing investedmoney Lockup period is considered to be zero for a fund that has no lock-up provision Subscriptionperiod is the time delay measured in quarters between investing in a fund and actually purchasingfund shares Constrained is an indicator variable that equals one if the fundrsquos strategy belongs toConvertible Arbitrage Fixed Income Arbitrage Emerging Markets and Event Driven strategiesand zero otherwise (Ding et al (2009))
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 50333 798 minus413 034 1014 3701Quarterly returns () 50333 256 minus091 211 566 905AUM ($ million) 50333 2108 110 393 1300 6370Age (quarters) 50333 1756 609 1319 2434 1532Volatility () 50333 533 300 441 804 243Management fees ( annualized) 2998 15 10 15 20 05Incentive fees () 2998 193 200 200 200 34HWM 2998 0694 0000 1000 1000 0461Leverage 2998 0685 0000 1000 1000 0464Open-to-public 2998 0191 0000 0000 0000 0394On-shore 2998 0230 0000 0000 0000 0421Total redemption period (quarters) 2998 1085 0556 0667 1333 1015Lockup period (quarters) 2998 0974 0000 0000 0000 2077Subscription period (quarters) 2998 0356 0333 0333 0333 0190Style
Constrained 2998 0303 0000 0000 1000 0460Constrained = 1
Convertible Arbitrage 2998 0034Emerging Markets 2998 0120Event Driven 2998 0096Fixed Income Arbitrage 2998 0052
Constrained = 0Equity Market Neutral 2998 0073Global Macro 2998 0078LongShort Equity Hedge 2998 0420Multi-Strategy 2998 0077Other 2998 0049
Indirect Incentives of Hedge Fund Managers 879
Our sample period is from January 1995 to December 2010 We focus onthe post-1994 period because TASS started reporting information on ldquodefunctrdquofunds only after 19945 We exclude managed futuresCTAs and funds-of-fundswhich have different incentive fees and are likely to have different inflow-performance relations from typical individual hedge funds We also excludeclosed-end hedge funds as subscriptions in these funds are only possible duringthe initial issuing period so future flows are not possible This initial filterleaves us with 4939 open-end hedge funds6
We drop funds for which TASS does not contain information on organizationalcharacteristics such as management fees incentive fees and HWM provisionsIn addition we consider only funds with an uninterrupted series of net assetvalues and returns so that we can calculate inflows Further we restrict oursample to funds with at least 12 consecutive monthly returns available duringthe sample period If a fund stops reporting returns and then resumes at afuture date we use only the first sequence of uninterrupted data Finally weexclude funds with an incentive fee of zero because there can be no direct payfor performance for these funds
We conduct the analysis using quarterly data because we wish to include onlylagged (not contemporaneous) returns in the flow-performance specificationsand to have a relatively short gap between returns and flows7 To construct aquarterly sample we drop all fund-calendar quarters that have return infor-mation only for a fraction of the period We also require that a sample fundhave a subscription period less than or equal to three months so that quarterlyinflows are not restricted This sample construction process leaves us with asample of 2998 funds (50333 fund-quarter observations)
Table I presents descriptive statistics Time-varying variables such as Quar-terly flow and Quarterly returns are measured at the fund-quarter level andother contractual characteristics such as Management fees and Incentive feesare measured at the fund level8 All time-varying variables except fund age(Age) are winsorized at the 1st and 99th percentiles to minimize the effect ofoutliers
5 Defunct funds include funds that are liquidated merged or restructured as well as those thatstopped reporting returns to TASS (Fung et al (2008))
6 Some funds are closed to new investors but unfortunately we do not know whether a particularfund is taking new money at any point in time so we cannot exclude funds on the basis of thispolicy Including closed funds causes us to understate the flow-performance relation for funds thatare not closed
7 Prior versions of this paper presented estimates using annual data with contemporaneousannual returns included in the flow-performance specifications (Lim Sensoy and Weisbach (2013))The estimates of indirect incentives using annual data are very close to those reported here
8 TASS provides information on fundsrsquo organizational characteristics as of the last available dateof fund data Like most previous studies we also assume that these organizational characteristicsdo not change throughout the life of the fund Agarwal Daniel and Naik (2009) argue that fundsrsquoorganizational characteristics are unlikely to change much over time based on their discussionswith practitioners which suggests that if a manager wants to impose new contractual terms it iseasier for him to start a new fund with different terms than to go through the legal complicationsof changing an existing contract
880 The Journal of Finance Rcopy
Mean Quarterly flow is 80 with a median of 03 so the distributionis highly skewed Mean Quarterly returns is 26 the median is 21 Aver-age fund size (AUM) is $2108 million The remaining variables reflect time-invariant contractual features Summary statistics on these characteristics arevery close to those reported in prior studies (eg Agarwal Daniel and Naik(2009) Baquero and Verbeek (2009) Aragon and Nanda (2012)) The variableManagement fees is the annual percentage of fund assets (AUM) received bythe manager as compensation and has a sample mean (median) of 15 (15)The variable Incentive fees is the percentage of profits above the HWM receivedby the manager as compensation and has a sample mean (median) of 193(20) Over two-thirds of our sample funds 694 have a HWM provision685 report that they use leverage 191 are open to public investors andabout a quarter are on-shore funds
The fee data consist of the fees that are currently publicly quoted by thefunds These data could potentially misrepresent the true fees relevant forour calculations for three reasons First funds sometimes provide fee reduc-tions to particular strategic investors that are not reflected in the database(Ramadorai and Streatfield (2011)) Although we cannot investigate this pos-sibility directly it is unlikely to have a major impact on our conclusions Forexample if the true incentive fee (management fee) averaged over all investorswere 10 lower than what is reported in the database our estimates of indirectincentives using the base GIR model would be overstated by about 1 (5) Asecond issue is that fees can change over time Agarwal and Ray (2012) andDueskar et al (2012) find that fee changes are infrequent and tend to reflectpast performance when they do occur so that fee increases follow good perfor-mance and fee decreases follow poor performance This effect is magnified inindirect incentives because good performance today leads not only to inflowsbut also to higher proportional fees on those inflows Third it is possible thatthere are fee changes unobservable to researchers To the extent that observ-able and unobservable fee changes are positively correlated this possibilityleads us to understate indirect incentives
We consider three variables that reflect potential restrictions on the be-havior of flows First we use Total redemption period defined as the sum ofthe notice period and the redemption period where the notice period is thetime the investor has to give notice to the fund about an intention to with-draw money from the fund and the redemption period is the time that thefund takes to return the money after the notice period is over The Lockupperiod is the minimum time that an investor has to wait before withdrawinginvested money The Subscription period is the time delay between investingin a fund and actually purchasing fund shares The mean total redemption pe-riod lock-up period and subscription period are 109 097 and 036 quartersrespectively
Indirect Incentives of Hedge Fund Managers 881
III Estimating the Sensitivity of Fund Inflows to Performance
To understand the impact of performance on fund flows we employ aBayesian learning framework that presumes investors are continually eval-uating managers trying to assess their abilities (see Berk and Green (2004)Chung et al (2012)) A fundrsquos performance provides information about the man-agerrsquos ability so an observation of performance will lead investors to updatetheir assessment of his ability and allocate more capital to a fund if their es-timate of the managerrsquos ability increases The magnitude of the updates andhence the sensitivity of inflows to performance should depend on the infor-mativeness of the signal relative to the precision of the prior estimate of thefund managerrsquos ability The sensitivity of inflows to performance should alsodepend on the extent to which ability can be scaled to replicate a fundrsquos returndistribution on new capital
Measuring the indirect incentives of hedge fund managers requires an es-timate of the relation between fund performance and future inflows A longliterature beginning with Ippolito (1992) estimates this relation to be rela-tively strong in the mutual fund industry9 It is also positive in the privateequity industry (see eg Kaplan and Schoar (2005)) However the results forhedge funds are less clear Goetzmann Ingersoll and Ross (2003) report a neg-ative and concave relation whereas other studies including Agarwal Danieland Naik (2003) Fung et al (2008) Baquero and Verbeek (2009) and Dinget al (2009) find a positive one
A Empirical Specification
We estimate the following specification
Flowit = β0 +11sumj=1
β0+ j Returnitminus j + γ Xtminus1 + λY + Fixed effects + εit (1)
where Flowit represents flows for fund i in quarter t10
Following the literature on flows to mutual funds (eg Sirri and Tufano(1998) Chevalier and Ellison (1999)) or hedge funds (eg Fung et al (2008)Agarwal Daniel and Naik (2009)) we compute quarterly flows of capital intoa fund as follows
Flowit = AUMit minus AUMitminus1(1 + Returnit
)AUMitminus1
(2)
where AUMit and AUMitminus1 are the AUM of fund i at the end of quarters t andtminus1 respectively and Returnit is the net-of-fee return for fund i in quarter t
9 See Brown Halow and Starks (1996) Sirri and Tufano (1998) Chevalier and Ellison (1999)Barclay Pearson and Weisbach (1998) Del Guercio and Tkac (2002) Bollen (2007) Huang Weiand Yan (2007) and Sensoy (2009)
10 We restrict our estimates to quarterly specifications but prior versions of this paper alsoinclude annual and monthly specifications with similar results to those reported below
882 The Journal of Finance Rcopy
This definition expresses flows as a fraction of beginning-of-period (end ofprior period) AUM which is a natural benchmark from the perspective of afund manager assessing his or her incentives going forward For instance theoption deltas that comprise direct incentives are defined in terms of beginning-of-period AUM An alternative approach to computing fund flows is to scalethe denominator of equation (2) by (1 + Returnit) which expresses flows as afraction of what end-of-period AUM would have been in the absence of flowsAlthough this alternative definition is less intuitive for our purpose it leads tosimilar estimates of indirect incentives
The vector X in equation (1) consists of time-varying fund characteristicsthat include lagged flows the natural logarithm of AUM for fund i at tminus1 thenatural logarithm of one plus fund irsquos age in quarters at tminus1 and fund volatil-ity over the previous 12 months11 The vector Y includes time-invariant fundcharacteristics that include Management fees Incentive fees Total redemptionperiod Lockup period Subscription period and a set of indicator variables thatequal one if the fund has an HWM provision uses leverage is open to publicinvestors and is an on-shore fund All specifications include fixed effects forthe nine styles listed in Table I interacted with calendar quarter fixed effectsThese time-by-style effects capture all shocks observed or unobserved that arecommon to funds of a given style in a given quarter including the returns topeer funds and inflows to other funds of the same style12 Reported standarderrors are robust to heteroskedasticity and account for double clustering byfund and time period13
B Estimates of the Flow-Performance Relation
We present estimates of the flow-performance relation in Panel A ofTable II In Column (1) we include returns in the 11 quarters prior to thecurrent quarter In Column (2) we add a number of fund-level controls In eachspecification the coefficients on returns are positive and statistically signif-icant in most cases and decline sharply over time so the coefficient on themost recent quarterrsquos return is the largest If we sum the coefficients on the 11prior quarters the sum in Column (1) is 144 and in Column (2) is 150 Thesecoefficients imply that a one-standard-deviation increase in returns (9) willlead to about a 13 increase in fund size
11 For young funds that have for example only one yearrsquos worth of return history we cannotcompute lagged returns and flows In such cases we ldquodummy outrdquo missing lagged variables toretain observations To do so we set missing values of lagged flows and returns to zero and includean indicator for missing values
12 Time-by-style fixed effects perform the same adjustment as a factor model regression underthe assumption that the factor loadings are the same for all funds of a given style within a giventime period
13 We focus on linear specifications Although there is some evidence of nonlinear effects in thedata (using splines quadratics etc) they are small in magnitude and have little impact on theestimates obtained with linear specifications
Indirect Incentives of Hedge Fund Managers 883
Table IIEstimates of Flow-Performance Sensitivity
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) using quarterly data The dependent variable is Quarterly flow as defined byequation (2) See Table I for definitions of all independent variables The sample period is from 1995to 2010 We only employ the funds with a subscription period less than or equal to three months toavoid quarterly inflows being restricted by subscription periods Equation (1) is augmented in PanelB to include interaction terms of the log of the fundrsquos age plus one with prior performance and inPanel C to include interaction terms of Constrained with performance variables All specificationsin Panel A and B include style fixed effects quarter fixed effects and quarter-by-style fixed effectsOnly quarter fixed effects are included in Panel C due to multicollinearity Standard errors aredouble clustered by fund and year and indicate statistical significance at the 1 5and 10 level respectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel A Base-case
Quarterly returns(tminus1) 0501 (0045) 0496 (0043)Quarterly returns(tminus2) 0314 (0039) 0263 (0036)Quarterly returns(tminus3) 0209 (0034) 0190 (0028)Quarterly returns(tminus4) 0117 (0035) 0111 (0030)Quarterly returns(tminus5) 0089 (0030) 0098 (0027)Quarterly returns(tminus6) 0054 (0033) 0066 (0030)Quarterly returns(tminus7) 0083 (0027) 0103 (0025)Quarterly returns(tminus8) 0044 (0022) 0060 (0022)Quarterly returns(tminus9) 0012 (0029) 0034 (0029)Quarterly returns(tminus10) minus0010 (0030) 0019 (0029)Quarterly returns(tminus11) 0025 (0027) 0056 (0026)Quarterly flow(tminus1) 0158 (0008)Log(Age + 1) minus0004 (0005)Log(AUM(tminus1)+1) minus0024 (0001)Volatility minus0141 (0037)Management fees 0211 (0386)Incentive fees 0005 (0053)HWM indicator 0020 (0004)Leveraged indicator 0000 (0004)Open-to-public indicator minus0002 (0005)On-shore indicator minus0032 (0005)Log(Total redemption period +1) 0034 (0007)Log(Lock-up period +1) minus0005 (0003)Log(Subscription period +1) minus0023 (0015)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0162 0194F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1940 (0000) 2077 (0000)
(Continued)
884 The Journal of Finance Rcopy
Table IImdashContinued
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel B Age interactions
Quarterly returns(tminus1)timesLog(Age in quarters+1) minus0319 (0035) minus0285 (0034)Quarterly returns(tminus2)timesLog(Age in quarters+1) minus0171 (0027) minus0125 (0025)Quarterly returns(tminus3)timesLog(Age in quarters+1) minus0051 (0036) minus0033 (0036)Quarterly returns(tminus4)timesLog(Age in quarters+1) minus0003 (0032) 0001 (0032)Quarterly returns(tminus5)timesLog(Age in quarters+1) minus0041 (0030) minus0042 (0029)Quarterly returns(tminus6)timesLog(Age in quarters+1) 0038 (0033) 0040 (0031)Quarterly returns(tminus7)timesLog(Age in quarters+1) 0013 (0033) 0007 (0032)Quarterly returns(tminus8)timesLog(Age in quarters+1) 0033 (0033) 0034 (0033)Quarterly returns(tminus9)timesLog(Age in quarters+1) 0003 (0040) minus0004 (0040)Quarterly returns(tminus10)timesLog(Age in quarters+1) 0132 (0038) 0133 (0035)Quarterly returns(tminus11)timesLog(Age in quarters+1) 0018 (0040) 0007 (0041)Quarterly returns(tminus1) 1325 (0117) 1233 (0113)Quarterly returns(tminus2) 0773 (0095) 0602 (0087)Quarterly returns(tminus3) 0354 (0114) 0287 (0110)Quarterly returns(tminus4) 0124 (0112) 0110 (0106)Quarterly returns(tminus5) 0200 (0100) 0215 (0092)Quarterly returns(tminus6) minus0067 (0112) minus0056 (0104)Quarterly returns(tminus7) 0033 (0106) 0079 (0099)Quarterly returns(tminus8) minus0066 (0107) minus0046 (0107)Quarterly returns(tminus9) 0004 (0128) 0053 (0127)Quarterly returns(tminus10) minus0416 (0122) minus0383 (0114)Quarterly returns(tminus11) minus0038 (0126) 0037 (0130)Log(Age + 1) minus0026 (0007) minus0000 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0168 0198F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1788 (0000) 1570 (0000)All age interaction terms 1214 (0000) 1084 (0000)
Panel C Scalability interactions
Quarterly returns(tminus1)timesConstrained minus0215 (0048) minus0215 (0047)Quarterly returns(tminus2)timesConstrained minus0096 (0043) minus0078 (0039)Quarterly returns(tminus3)timesConstrained minus0102 (0038) minus0107 (0037)Quarterly returns(tminus4)timesConstrained 0013 (0034) 0013 (0036)Quarterly returns(tminus5)timesConstrained minus0009 (0032) minus0022 (0032)Quarterly returns(tminus6)timesConstrained 0026 (0034) 0013 (0034)Quarterly returns(tminus7)timesConstrained minus0044 (0036) minus0057 (0036)Quarterly returns(tminus8)timesConstrained 0032 (0037) 0033 (0036)Quarterly returns(tminus9)timesConstrained minus0085 (0030) minus0090 (0031)Quarterly returns(tminus10)timesConstrained minus0023 (0036) minus0007 (0036)Quarterly returns(tminus11)timesConstrained minus0007 (0037) minus0004 (0038)
(Continued)
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 877
which leverage is zero so all four sets of estimates can be viewed as comingfrom variations of the LWY model
Using different models to estimate indirect incentives allows us to gaugethe importance of different factors especially the importance of futureperformance-based fund flows and endogenous portfolio choice in determin-ing the fraction of a fundrsquos value that will ultimately accrue to managers infuture fees3 Details on all four models our choices of model parameters andthe associated calculations are described in Appendix B
For each of the models above we estimate the present value of the total(management plus incentive) future fees that the manager earns on an extradollar of AUM To calculate indirect pay for performance we multiply thispresent value by an estimate of the number of incremental dollars of AUMthat result from a one-percentage-point increase in returns to investors Thelatter increase consists of two parts namely the mechanical increase in thevalue of existing investorsrsquo stakes plus incremental inflows of new investmentIn this way the models for a managerrsquos fee value together with our estimatesof the flow-performance relation facing hedge fund managers provide an esti-mate of the present value of the incremental future revenue that the hedgefund manager expects to earn as a result of a one-percentage-point improve-ment in current net returns4 This calculation results in an incentive measuredenominated in dollars per percentage point change in returns We also con-duct the same calculation for a one-dollar improvement in investor return toobtain a measure in terms of dollars in managerial wealth per dollar returnedto shareholders Both of these measures are commonly used in the literatureon manager incentives
Because the GIR and LWY models estimate present values no further ad-justment for the riskiness of future income is required Also the estimates donot require that the manager continue to manage the fund in the future underthe assumption that the present value of the managerrsquos claims to future feeincome can be monetized when the manager departs
II Hedge Fund Data
Our data come from the TASS database which covers about 40 of thehedge fund universe (Agarwal Daniel and Naik (2009)) Summary statisticsfor key sample fund characteristics reported in Table I are very close to thosefor the sample considered by Agarwal Daniel and Naik (2009) who mergeand consolidate four major databases (CISDM HFR MSCI and TASS) Wetherefore conclude that our sample of hedge funds is representative of thehedge fund universe
3 In addition to GIR and LWY there have been a number of other attempts to value managersrsquoclaims to hedge funds Important contributions include Panageas and Westerfield (2009) Drechsler(2014) and Guasoni and Oblόj (2013)
4 Net returns can be improved either through increased gross returns or decreased costs borneby the fund such as financing costs security lending fees and settlement charges The incentiveswe measure are therefore incentives to achieve both
878 The Journal of Finance Rcopy
Table ISummary Statistics
This table presents descriptive statistics for the sample of 2998 funds during the 1995 to 2010sample period Time-varying variables are reported at the fund-quarter level and other time-invariant contractual characteristics are reported at the fund level Quarterly flow is the differencebetween the reported quarter-end AUM and the quarter-beginning AUM times (1 + Quarterlyreturns) scaled by quarter-beginning AUM Quarterly returns are the reported quarterly net-of-fee returns AUM is assets under management in millions measured at the end of each quarterAge is the number of quarters from the fundrsquos inception date measured at the beginning of eachquarter Volatility is the standard deviation of monthly (net-of-fee) returns over the 12 monthsprior to each quarter multiplied by the square root of three to arrive at a quarterly figure Alltime-varying variables except Age are winsorized at the 1st and 99th percentiles Management feesare the percentage of the assets charged by the fund as regular fees in annual terms Incentive feesare the percentage of positive profits received by the manager as performance incentives HWMis an indicator variable that equals one if the fund has a HWM provision and zero otherwiseLeverage is an indicator variable that equals one if the fund uses leverage and zero otherwiseOpen-to-public is an indicator variable that equals one if the fund allows public investment andzero otherwise On-shore is an indicator variable that equals one if the fund is domiciled in theUnited States and zero otherwise Total redemption period is the sum of the notice period and theredemption period measured in quarters where the notice period is the time the investor has togive notice to the fund about an intention to withdraw money from the fund and the redemptionperiod is the time that the fund takes to return the money after the notice period is over Lockupperiod is the minimum time in quarters that an investor has to wait before withdrawing investedmoney Lockup period is considered to be zero for a fund that has no lock-up provision Subscriptionperiod is the time delay measured in quarters between investing in a fund and actually purchasingfund shares Constrained is an indicator variable that equals one if the fundrsquos strategy belongs toConvertible Arbitrage Fixed Income Arbitrage Emerging Markets and Event Driven strategiesand zero otherwise (Ding et al (2009))
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 50333 798 minus413 034 1014 3701Quarterly returns () 50333 256 minus091 211 566 905AUM ($ million) 50333 2108 110 393 1300 6370Age (quarters) 50333 1756 609 1319 2434 1532Volatility () 50333 533 300 441 804 243Management fees ( annualized) 2998 15 10 15 20 05Incentive fees () 2998 193 200 200 200 34HWM 2998 0694 0000 1000 1000 0461Leverage 2998 0685 0000 1000 1000 0464Open-to-public 2998 0191 0000 0000 0000 0394On-shore 2998 0230 0000 0000 0000 0421Total redemption period (quarters) 2998 1085 0556 0667 1333 1015Lockup period (quarters) 2998 0974 0000 0000 0000 2077Subscription period (quarters) 2998 0356 0333 0333 0333 0190Style
Constrained 2998 0303 0000 0000 1000 0460Constrained = 1
Convertible Arbitrage 2998 0034Emerging Markets 2998 0120Event Driven 2998 0096Fixed Income Arbitrage 2998 0052
Constrained = 0Equity Market Neutral 2998 0073Global Macro 2998 0078LongShort Equity Hedge 2998 0420Multi-Strategy 2998 0077Other 2998 0049
Indirect Incentives of Hedge Fund Managers 879
Our sample period is from January 1995 to December 2010 We focus onthe post-1994 period because TASS started reporting information on ldquodefunctrdquofunds only after 19945 We exclude managed futuresCTAs and funds-of-fundswhich have different incentive fees and are likely to have different inflow-performance relations from typical individual hedge funds We also excludeclosed-end hedge funds as subscriptions in these funds are only possible duringthe initial issuing period so future flows are not possible This initial filterleaves us with 4939 open-end hedge funds6
We drop funds for which TASS does not contain information on organizationalcharacteristics such as management fees incentive fees and HWM provisionsIn addition we consider only funds with an uninterrupted series of net assetvalues and returns so that we can calculate inflows Further we restrict oursample to funds with at least 12 consecutive monthly returns available duringthe sample period If a fund stops reporting returns and then resumes at afuture date we use only the first sequence of uninterrupted data Finally weexclude funds with an incentive fee of zero because there can be no direct payfor performance for these funds
We conduct the analysis using quarterly data because we wish to include onlylagged (not contemporaneous) returns in the flow-performance specificationsand to have a relatively short gap between returns and flows7 To construct aquarterly sample we drop all fund-calendar quarters that have return infor-mation only for a fraction of the period We also require that a sample fundhave a subscription period less than or equal to three months so that quarterlyinflows are not restricted This sample construction process leaves us with asample of 2998 funds (50333 fund-quarter observations)
Table I presents descriptive statistics Time-varying variables such as Quar-terly flow and Quarterly returns are measured at the fund-quarter level andother contractual characteristics such as Management fees and Incentive feesare measured at the fund level8 All time-varying variables except fund age(Age) are winsorized at the 1st and 99th percentiles to minimize the effect ofoutliers
5 Defunct funds include funds that are liquidated merged or restructured as well as those thatstopped reporting returns to TASS (Fung et al (2008))
6 Some funds are closed to new investors but unfortunately we do not know whether a particularfund is taking new money at any point in time so we cannot exclude funds on the basis of thispolicy Including closed funds causes us to understate the flow-performance relation for funds thatare not closed
7 Prior versions of this paper presented estimates using annual data with contemporaneousannual returns included in the flow-performance specifications (Lim Sensoy and Weisbach (2013))The estimates of indirect incentives using annual data are very close to those reported here
8 TASS provides information on fundsrsquo organizational characteristics as of the last available dateof fund data Like most previous studies we also assume that these organizational characteristicsdo not change throughout the life of the fund Agarwal Daniel and Naik (2009) argue that fundsrsquoorganizational characteristics are unlikely to change much over time based on their discussionswith practitioners which suggests that if a manager wants to impose new contractual terms it iseasier for him to start a new fund with different terms than to go through the legal complicationsof changing an existing contract
880 The Journal of Finance Rcopy
Mean Quarterly flow is 80 with a median of 03 so the distributionis highly skewed Mean Quarterly returns is 26 the median is 21 Aver-age fund size (AUM) is $2108 million The remaining variables reflect time-invariant contractual features Summary statistics on these characteristics arevery close to those reported in prior studies (eg Agarwal Daniel and Naik(2009) Baquero and Verbeek (2009) Aragon and Nanda (2012)) The variableManagement fees is the annual percentage of fund assets (AUM) received bythe manager as compensation and has a sample mean (median) of 15 (15)The variable Incentive fees is the percentage of profits above the HWM receivedby the manager as compensation and has a sample mean (median) of 193(20) Over two-thirds of our sample funds 694 have a HWM provision685 report that they use leverage 191 are open to public investors andabout a quarter are on-shore funds
The fee data consist of the fees that are currently publicly quoted by thefunds These data could potentially misrepresent the true fees relevant forour calculations for three reasons First funds sometimes provide fee reduc-tions to particular strategic investors that are not reflected in the database(Ramadorai and Streatfield (2011)) Although we cannot investigate this pos-sibility directly it is unlikely to have a major impact on our conclusions Forexample if the true incentive fee (management fee) averaged over all investorswere 10 lower than what is reported in the database our estimates of indirectincentives using the base GIR model would be overstated by about 1 (5) Asecond issue is that fees can change over time Agarwal and Ray (2012) andDueskar et al (2012) find that fee changes are infrequent and tend to reflectpast performance when they do occur so that fee increases follow good perfor-mance and fee decreases follow poor performance This effect is magnified inindirect incentives because good performance today leads not only to inflowsbut also to higher proportional fees on those inflows Third it is possible thatthere are fee changes unobservable to researchers To the extent that observ-able and unobservable fee changes are positively correlated this possibilityleads us to understate indirect incentives
We consider three variables that reflect potential restrictions on the be-havior of flows First we use Total redemption period defined as the sum ofthe notice period and the redemption period where the notice period is thetime the investor has to give notice to the fund about an intention to with-draw money from the fund and the redemption period is the time that thefund takes to return the money after the notice period is over The Lockupperiod is the minimum time that an investor has to wait before withdrawinginvested money The Subscription period is the time delay between investingin a fund and actually purchasing fund shares The mean total redemption pe-riod lock-up period and subscription period are 109 097 and 036 quartersrespectively
Indirect Incentives of Hedge Fund Managers 881
III Estimating the Sensitivity of Fund Inflows to Performance
To understand the impact of performance on fund flows we employ aBayesian learning framework that presumes investors are continually eval-uating managers trying to assess their abilities (see Berk and Green (2004)Chung et al (2012)) A fundrsquos performance provides information about the man-agerrsquos ability so an observation of performance will lead investors to updatetheir assessment of his ability and allocate more capital to a fund if their es-timate of the managerrsquos ability increases The magnitude of the updates andhence the sensitivity of inflows to performance should depend on the infor-mativeness of the signal relative to the precision of the prior estimate of thefund managerrsquos ability The sensitivity of inflows to performance should alsodepend on the extent to which ability can be scaled to replicate a fundrsquos returndistribution on new capital
Measuring the indirect incentives of hedge fund managers requires an es-timate of the relation between fund performance and future inflows A longliterature beginning with Ippolito (1992) estimates this relation to be rela-tively strong in the mutual fund industry9 It is also positive in the privateequity industry (see eg Kaplan and Schoar (2005)) However the results forhedge funds are less clear Goetzmann Ingersoll and Ross (2003) report a neg-ative and concave relation whereas other studies including Agarwal Danieland Naik (2003) Fung et al (2008) Baquero and Verbeek (2009) and Dinget al (2009) find a positive one
A Empirical Specification
We estimate the following specification
Flowit = β0 +11sumj=1
β0+ j Returnitminus j + γ Xtminus1 + λY + Fixed effects + εit (1)
where Flowit represents flows for fund i in quarter t10
Following the literature on flows to mutual funds (eg Sirri and Tufano(1998) Chevalier and Ellison (1999)) or hedge funds (eg Fung et al (2008)Agarwal Daniel and Naik (2009)) we compute quarterly flows of capital intoa fund as follows
Flowit = AUMit minus AUMitminus1(1 + Returnit
)AUMitminus1
(2)
where AUMit and AUMitminus1 are the AUM of fund i at the end of quarters t andtminus1 respectively and Returnit is the net-of-fee return for fund i in quarter t
9 See Brown Halow and Starks (1996) Sirri and Tufano (1998) Chevalier and Ellison (1999)Barclay Pearson and Weisbach (1998) Del Guercio and Tkac (2002) Bollen (2007) Huang Weiand Yan (2007) and Sensoy (2009)
10 We restrict our estimates to quarterly specifications but prior versions of this paper alsoinclude annual and monthly specifications with similar results to those reported below
882 The Journal of Finance Rcopy
This definition expresses flows as a fraction of beginning-of-period (end ofprior period) AUM which is a natural benchmark from the perspective of afund manager assessing his or her incentives going forward For instance theoption deltas that comprise direct incentives are defined in terms of beginning-of-period AUM An alternative approach to computing fund flows is to scalethe denominator of equation (2) by (1 + Returnit) which expresses flows as afraction of what end-of-period AUM would have been in the absence of flowsAlthough this alternative definition is less intuitive for our purpose it leads tosimilar estimates of indirect incentives
The vector X in equation (1) consists of time-varying fund characteristicsthat include lagged flows the natural logarithm of AUM for fund i at tminus1 thenatural logarithm of one plus fund irsquos age in quarters at tminus1 and fund volatil-ity over the previous 12 months11 The vector Y includes time-invariant fundcharacteristics that include Management fees Incentive fees Total redemptionperiod Lockup period Subscription period and a set of indicator variables thatequal one if the fund has an HWM provision uses leverage is open to publicinvestors and is an on-shore fund All specifications include fixed effects forthe nine styles listed in Table I interacted with calendar quarter fixed effectsThese time-by-style effects capture all shocks observed or unobserved that arecommon to funds of a given style in a given quarter including the returns topeer funds and inflows to other funds of the same style12 Reported standarderrors are robust to heteroskedasticity and account for double clustering byfund and time period13
B Estimates of the Flow-Performance Relation
We present estimates of the flow-performance relation in Panel A ofTable II In Column (1) we include returns in the 11 quarters prior to thecurrent quarter In Column (2) we add a number of fund-level controls In eachspecification the coefficients on returns are positive and statistically signif-icant in most cases and decline sharply over time so the coefficient on themost recent quarterrsquos return is the largest If we sum the coefficients on the 11prior quarters the sum in Column (1) is 144 and in Column (2) is 150 Thesecoefficients imply that a one-standard-deviation increase in returns (9) willlead to about a 13 increase in fund size
11 For young funds that have for example only one yearrsquos worth of return history we cannotcompute lagged returns and flows In such cases we ldquodummy outrdquo missing lagged variables toretain observations To do so we set missing values of lagged flows and returns to zero and includean indicator for missing values
12 Time-by-style fixed effects perform the same adjustment as a factor model regression underthe assumption that the factor loadings are the same for all funds of a given style within a giventime period
13 We focus on linear specifications Although there is some evidence of nonlinear effects in thedata (using splines quadratics etc) they are small in magnitude and have little impact on theestimates obtained with linear specifications
Indirect Incentives of Hedge Fund Managers 883
Table IIEstimates of Flow-Performance Sensitivity
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) using quarterly data The dependent variable is Quarterly flow as defined byequation (2) See Table I for definitions of all independent variables The sample period is from 1995to 2010 We only employ the funds with a subscription period less than or equal to three months toavoid quarterly inflows being restricted by subscription periods Equation (1) is augmented in PanelB to include interaction terms of the log of the fundrsquos age plus one with prior performance and inPanel C to include interaction terms of Constrained with performance variables All specificationsin Panel A and B include style fixed effects quarter fixed effects and quarter-by-style fixed effectsOnly quarter fixed effects are included in Panel C due to multicollinearity Standard errors aredouble clustered by fund and year and indicate statistical significance at the 1 5and 10 level respectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel A Base-case
Quarterly returns(tminus1) 0501 (0045) 0496 (0043)Quarterly returns(tminus2) 0314 (0039) 0263 (0036)Quarterly returns(tminus3) 0209 (0034) 0190 (0028)Quarterly returns(tminus4) 0117 (0035) 0111 (0030)Quarterly returns(tminus5) 0089 (0030) 0098 (0027)Quarterly returns(tminus6) 0054 (0033) 0066 (0030)Quarterly returns(tminus7) 0083 (0027) 0103 (0025)Quarterly returns(tminus8) 0044 (0022) 0060 (0022)Quarterly returns(tminus9) 0012 (0029) 0034 (0029)Quarterly returns(tminus10) minus0010 (0030) 0019 (0029)Quarterly returns(tminus11) 0025 (0027) 0056 (0026)Quarterly flow(tminus1) 0158 (0008)Log(Age + 1) minus0004 (0005)Log(AUM(tminus1)+1) minus0024 (0001)Volatility minus0141 (0037)Management fees 0211 (0386)Incentive fees 0005 (0053)HWM indicator 0020 (0004)Leveraged indicator 0000 (0004)Open-to-public indicator minus0002 (0005)On-shore indicator minus0032 (0005)Log(Total redemption period +1) 0034 (0007)Log(Lock-up period +1) minus0005 (0003)Log(Subscription period +1) minus0023 (0015)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0162 0194F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1940 (0000) 2077 (0000)
(Continued)
884 The Journal of Finance Rcopy
Table IImdashContinued
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel B Age interactions
Quarterly returns(tminus1)timesLog(Age in quarters+1) minus0319 (0035) minus0285 (0034)Quarterly returns(tminus2)timesLog(Age in quarters+1) minus0171 (0027) minus0125 (0025)Quarterly returns(tminus3)timesLog(Age in quarters+1) minus0051 (0036) minus0033 (0036)Quarterly returns(tminus4)timesLog(Age in quarters+1) minus0003 (0032) 0001 (0032)Quarterly returns(tminus5)timesLog(Age in quarters+1) minus0041 (0030) minus0042 (0029)Quarterly returns(tminus6)timesLog(Age in quarters+1) 0038 (0033) 0040 (0031)Quarterly returns(tminus7)timesLog(Age in quarters+1) 0013 (0033) 0007 (0032)Quarterly returns(tminus8)timesLog(Age in quarters+1) 0033 (0033) 0034 (0033)Quarterly returns(tminus9)timesLog(Age in quarters+1) 0003 (0040) minus0004 (0040)Quarterly returns(tminus10)timesLog(Age in quarters+1) 0132 (0038) 0133 (0035)Quarterly returns(tminus11)timesLog(Age in quarters+1) 0018 (0040) 0007 (0041)Quarterly returns(tminus1) 1325 (0117) 1233 (0113)Quarterly returns(tminus2) 0773 (0095) 0602 (0087)Quarterly returns(tminus3) 0354 (0114) 0287 (0110)Quarterly returns(tminus4) 0124 (0112) 0110 (0106)Quarterly returns(tminus5) 0200 (0100) 0215 (0092)Quarterly returns(tminus6) minus0067 (0112) minus0056 (0104)Quarterly returns(tminus7) 0033 (0106) 0079 (0099)Quarterly returns(tminus8) minus0066 (0107) minus0046 (0107)Quarterly returns(tminus9) 0004 (0128) 0053 (0127)Quarterly returns(tminus10) minus0416 (0122) minus0383 (0114)Quarterly returns(tminus11) minus0038 (0126) 0037 (0130)Log(Age + 1) minus0026 (0007) minus0000 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0168 0198F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1788 (0000) 1570 (0000)All age interaction terms 1214 (0000) 1084 (0000)
Panel C Scalability interactions
Quarterly returns(tminus1)timesConstrained minus0215 (0048) minus0215 (0047)Quarterly returns(tminus2)timesConstrained minus0096 (0043) minus0078 (0039)Quarterly returns(tminus3)timesConstrained minus0102 (0038) minus0107 (0037)Quarterly returns(tminus4)timesConstrained 0013 (0034) 0013 (0036)Quarterly returns(tminus5)timesConstrained minus0009 (0032) minus0022 (0032)Quarterly returns(tminus6)timesConstrained 0026 (0034) 0013 (0034)Quarterly returns(tminus7)timesConstrained minus0044 (0036) minus0057 (0036)Quarterly returns(tminus8)timesConstrained 0032 (0037) 0033 (0036)Quarterly returns(tminus9)timesConstrained minus0085 (0030) minus0090 (0031)Quarterly returns(tminus10)timesConstrained minus0023 (0036) minus0007 (0036)Quarterly returns(tminus11)timesConstrained minus0007 (0037) minus0004 (0038)
(Continued)
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
878 The Journal of Finance Rcopy
Table ISummary Statistics
This table presents descriptive statistics for the sample of 2998 funds during the 1995 to 2010sample period Time-varying variables are reported at the fund-quarter level and other time-invariant contractual characteristics are reported at the fund level Quarterly flow is the differencebetween the reported quarter-end AUM and the quarter-beginning AUM times (1 + Quarterlyreturns) scaled by quarter-beginning AUM Quarterly returns are the reported quarterly net-of-fee returns AUM is assets under management in millions measured at the end of each quarterAge is the number of quarters from the fundrsquos inception date measured at the beginning of eachquarter Volatility is the standard deviation of monthly (net-of-fee) returns over the 12 monthsprior to each quarter multiplied by the square root of three to arrive at a quarterly figure Alltime-varying variables except Age are winsorized at the 1st and 99th percentiles Management feesare the percentage of the assets charged by the fund as regular fees in annual terms Incentive feesare the percentage of positive profits received by the manager as performance incentives HWMis an indicator variable that equals one if the fund has a HWM provision and zero otherwiseLeverage is an indicator variable that equals one if the fund uses leverage and zero otherwiseOpen-to-public is an indicator variable that equals one if the fund allows public investment andzero otherwise On-shore is an indicator variable that equals one if the fund is domiciled in theUnited States and zero otherwise Total redemption period is the sum of the notice period and theredemption period measured in quarters where the notice period is the time the investor has togive notice to the fund about an intention to withdraw money from the fund and the redemptionperiod is the time that the fund takes to return the money after the notice period is over Lockupperiod is the minimum time in quarters that an investor has to wait before withdrawing investedmoney Lockup period is considered to be zero for a fund that has no lock-up provision Subscriptionperiod is the time delay measured in quarters between investing in a fund and actually purchasingfund shares Constrained is an indicator variable that equals one if the fundrsquos strategy belongs toConvertible Arbitrage Fixed Income Arbitrage Emerging Markets and Event Driven strategiesand zero otherwise (Ding et al (2009))
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 50333 798 minus413 034 1014 3701Quarterly returns () 50333 256 minus091 211 566 905AUM ($ million) 50333 2108 110 393 1300 6370Age (quarters) 50333 1756 609 1319 2434 1532Volatility () 50333 533 300 441 804 243Management fees ( annualized) 2998 15 10 15 20 05Incentive fees () 2998 193 200 200 200 34HWM 2998 0694 0000 1000 1000 0461Leverage 2998 0685 0000 1000 1000 0464Open-to-public 2998 0191 0000 0000 0000 0394On-shore 2998 0230 0000 0000 0000 0421Total redemption period (quarters) 2998 1085 0556 0667 1333 1015Lockup period (quarters) 2998 0974 0000 0000 0000 2077Subscription period (quarters) 2998 0356 0333 0333 0333 0190Style
Constrained 2998 0303 0000 0000 1000 0460Constrained = 1
Convertible Arbitrage 2998 0034Emerging Markets 2998 0120Event Driven 2998 0096Fixed Income Arbitrage 2998 0052
Constrained = 0Equity Market Neutral 2998 0073Global Macro 2998 0078LongShort Equity Hedge 2998 0420Multi-Strategy 2998 0077Other 2998 0049
Indirect Incentives of Hedge Fund Managers 879
Our sample period is from January 1995 to December 2010 We focus onthe post-1994 period because TASS started reporting information on ldquodefunctrdquofunds only after 19945 We exclude managed futuresCTAs and funds-of-fundswhich have different incentive fees and are likely to have different inflow-performance relations from typical individual hedge funds We also excludeclosed-end hedge funds as subscriptions in these funds are only possible duringthe initial issuing period so future flows are not possible This initial filterleaves us with 4939 open-end hedge funds6
We drop funds for which TASS does not contain information on organizationalcharacteristics such as management fees incentive fees and HWM provisionsIn addition we consider only funds with an uninterrupted series of net assetvalues and returns so that we can calculate inflows Further we restrict oursample to funds with at least 12 consecutive monthly returns available duringthe sample period If a fund stops reporting returns and then resumes at afuture date we use only the first sequence of uninterrupted data Finally weexclude funds with an incentive fee of zero because there can be no direct payfor performance for these funds
We conduct the analysis using quarterly data because we wish to include onlylagged (not contemporaneous) returns in the flow-performance specificationsand to have a relatively short gap between returns and flows7 To construct aquarterly sample we drop all fund-calendar quarters that have return infor-mation only for a fraction of the period We also require that a sample fundhave a subscription period less than or equal to three months so that quarterlyinflows are not restricted This sample construction process leaves us with asample of 2998 funds (50333 fund-quarter observations)
Table I presents descriptive statistics Time-varying variables such as Quar-terly flow and Quarterly returns are measured at the fund-quarter level andother contractual characteristics such as Management fees and Incentive feesare measured at the fund level8 All time-varying variables except fund age(Age) are winsorized at the 1st and 99th percentiles to minimize the effect ofoutliers
5 Defunct funds include funds that are liquidated merged or restructured as well as those thatstopped reporting returns to TASS (Fung et al (2008))
6 Some funds are closed to new investors but unfortunately we do not know whether a particularfund is taking new money at any point in time so we cannot exclude funds on the basis of thispolicy Including closed funds causes us to understate the flow-performance relation for funds thatare not closed
7 Prior versions of this paper presented estimates using annual data with contemporaneousannual returns included in the flow-performance specifications (Lim Sensoy and Weisbach (2013))The estimates of indirect incentives using annual data are very close to those reported here
8 TASS provides information on fundsrsquo organizational characteristics as of the last available dateof fund data Like most previous studies we also assume that these organizational characteristicsdo not change throughout the life of the fund Agarwal Daniel and Naik (2009) argue that fundsrsquoorganizational characteristics are unlikely to change much over time based on their discussionswith practitioners which suggests that if a manager wants to impose new contractual terms it iseasier for him to start a new fund with different terms than to go through the legal complicationsof changing an existing contract
880 The Journal of Finance Rcopy
Mean Quarterly flow is 80 with a median of 03 so the distributionis highly skewed Mean Quarterly returns is 26 the median is 21 Aver-age fund size (AUM) is $2108 million The remaining variables reflect time-invariant contractual features Summary statistics on these characteristics arevery close to those reported in prior studies (eg Agarwal Daniel and Naik(2009) Baquero and Verbeek (2009) Aragon and Nanda (2012)) The variableManagement fees is the annual percentage of fund assets (AUM) received bythe manager as compensation and has a sample mean (median) of 15 (15)The variable Incentive fees is the percentage of profits above the HWM receivedby the manager as compensation and has a sample mean (median) of 193(20) Over two-thirds of our sample funds 694 have a HWM provision685 report that they use leverage 191 are open to public investors andabout a quarter are on-shore funds
The fee data consist of the fees that are currently publicly quoted by thefunds These data could potentially misrepresent the true fees relevant forour calculations for three reasons First funds sometimes provide fee reduc-tions to particular strategic investors that are not reflected in the database(Ramadorai and Streatfield (2011)) Although we cannot investigate this pos-sibility directly it is unlikely to have a major impact on our conclusions Forexample if the true incentive fee (management fee) averaged over all investorswere 10 lower than what is reported in the database our estimates of indirectincentives using the base GIR model would be overstated by about 1 (5) Asecond issue is that fees can change over time Agarwal and Ray (2012) andDueskar et al (2012) find that fee changes are infrequent and tend to reflectpast performance when they do occur so that fee increases follow good perfor-mance and fee decreases follow poor performance This effect is magnified inindirect incentives because good performance today leads not only to inflowsbut also to higher proportional fees on those inflows Third it is possible thatthere are fee changes unobservable to researchers To the extent that observ-able and unobservable fee changes are positively correlated this possibilityleads us to understate indirect incentives
We consider three variables that reflect potential restrictions on the be-havior of flows First we use Total redemption period defined as the sum ofthe notice period and the redemption period where the notice period is thetime the investor has to give notice to the fund about an intention to with-draw money from the fund and the redemption period is the time that thefund takes to return the money after the notice period is over The Lockupperiod is the minimum time that an investor has to wait before withdrawinginvested money The Subscription period is the time delay between investingin a fund and actually purchasing fund shares The mean total redemption pe-riod lock-up period and subscription period are 109 097 and 036 quartersrespectively
Indirect Incentives of Hedge Fund Managers 881
III Estimating the Sensitivity of Fund Inflows to Performance
To understand the impact of performance on fund flows we employ aBayesian learning framework that presumes investors are continually eval-uating managers trying to assess their abilities (see Berk and Green (2004)Chung et al (2012)) A fundrsquos performance provides information about the man-agerrsquos ability so an observation of performance will lead investors to updatetheir assessment of his ability and allocate more capital to a fund if their es-timate of the managerrsquos ability increases The magnitude of the updates andhence the sensitivity of inflows to performance should depend on the infor-mativeness of the signal relative to the precision of the prior estimate of thefund managerrsquos ability The sensitivity of inflows to performance should alsodepend on the extent to which ability can be scaled to replicate a fundrsquos returndistribution on new capital
Measuring the indirect incentives of hedge fund managers requires an es-timate of the relation between fund performance and future inflows A longliterature beginning with Ippolito (1992) estimates this relation to be rela-tively strong in the mutual fund industry9 It is also positive in the privateequity industry (see eg Kaplan and Schoar (2005)) However the results forhedge funds are less clear Goetzmann Ingersoll and Ross (2003) report a neg-ative and concave relation whereas other studies including Agarwal Danieland Naik (2003) Fung et al (2008) Baquero and Verbeek (2009) and Dinget al (2009) find a positive one
A Empirical Specification
We estimate the following specification
Flowit = β0 +11sumj=1
β0+ j Returnitminus j + γ Xtminus1 + λY + Fixed effects + εit (1)
where Flowit represents flows for fund i in quarter t10
Following the literature on flows to mutual funds (eg Sirri and Tufano(1998) Chevalier and Ellison (1999)) or hedge funds (eg Fung et al (2008)Agarwal Daniel and Naik (2009)) we compute quarterly flows of capital intoa fund as follows
Flowit = AUMit minus AUMitminus1(1 + Returnit
)AUMitminus1
(2)
where AUMit and AUMitminus1 are the AUM of fund i at the end of quarters t andtminus1 respectively and Returnit is the net-of-fee return for fund i in quarter t
9 See Brown Halow and Starks (1996) Sirri and Tufano (1998) Chevalier and Ellison (1999)Barclay Pearson and Weisbach (1998) Del Guercio and Tkac (2002) Bollen (2007) Huang Weiand Yan (2007) and Sensoy (2009)
10 We restrict our estimates to quarterly specifications but prior versions of this paper alsoinclude annual and monthly specifications with similar results to those reported below
882 The Journal of Finance Rcopy
This definition expresses flows as a fraction of beginning-of-period (end ofprior period) AUM which is a natural benchmark from the perspective of afund manager assessing his or her incentives going forward For instance theoption deltas that comprise direct incentives are defined in terms of beginning-of-period AUM An alternative approach to computing fund flows is to scalethe denominator of equation (2) by (1 + Returnit) which expresses flows as afraction of what end-of-period AUM would have been in the absence of flowsAlthough this alternative definition is less intuitive for our purpose it leads tosimilar estimates of indirect incentives
The vector X in equation (1) consists of time-varying fund characteristicsthat include lagged flows the natural logarithm of AUM for fund i at tminus1 thenatural logarithm of one plus fund irsquos age in quarters at tminus1 and fund volatil-ity over the previous 12 months11 The vector Y includes time-invariant fundcharacteristics that include Management fees Incentive fees Total redemptionperiod Lockup period Subscription period and a set of indicator variables thatequal one if the fund has an HWM provision uses leverage is open to publicinvestors and is an on-shore fund All specifications include fixed effects forthe nine styles listed in Table I interacted with calendar quarter fixed effectsThese time-by-style effects capture all shocks observed or unobserved that arecommon to funds of a given style in a given quarter including the returns topeer funds and inflows to other funds of the same style12 Reported standarderrors are robust to heteroskedasticity and account for double clustering byfund and time period13
B Estimates of the Flow-Performance Relation
We present estimates of the flow-performance relation in Panel A ofTable II In Column (1) we include returns in the 11 quarters prior to thecurrent quarter In Column (2) we add a number of fund-level controls In eachspecification the coefficients on returns are positive and statistically signif-icant in most cases and decline sharply over time so the coefficient on themost recent quarterrsquos return is the largest If we sum the coefficients on the 11prior quarters the sum in Column (1) is 144 and in Column (2) is 150 Thesecoefficients imply that a one-standard-deviation increase in returns (9) willlead to about a 13 increase in fund size
11 For young funds that have for example only one yearrsquos worth of return history we cannotcompute lagged returns and flows In such cases we ldquodummy outrdquo missing lagged variables toretain observations To do so we set missing values of lagged flows and returns to zero and includean indicator for missing values
12 Time-by-style fixed effects perform the same adjustment as a factor model regression underthe assumption that the factor loadings are the same for all funds of a given style within a giventime period
13 We focus on linear specifications Although there is some evidence of nonlinear effects in thedata (using splines quadratics etc) they are small in magnitude and have little impact on theestimates obtained with linear specifications
Indirect Incentives of Hedge Fund Managers 883
Table IIEstimates of Flow-Performance Sensitivity
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) using quarterly data The dependent variable is Quarterly flow as defined byequation (2) See Table I for definitions of all independent variables The sample period is from 1995to 2010 We only employ the funds with a subscription period less than or equal to three months toavoid quarterly inflows being restricted by subscription periods Equation (1) is augmented in PanelB to include interaction terms of the log of the fundrsquos age plus one with prior performance and inPanel C to include interaction terms of Constrained with performance variables All specificationsin Panel A and B include style fixed effects quarter fixed effects and quarter-by-style fixed effectsOnly quarter fixed effects are included in Panel C due to multicollinearity Standard errors aredouble clustered by fund and year and indicate statistical significance at the 1 5and 10 level respectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel A Base-case
Quarterly returns(tminus1) 0501 (0045) 0496 (0043)Quarterly returns(tminus2) 0314 (0039) 0263 (0036)Quarterly returns(tminus3) 0209 (0034) 0190 (0028)Quarterly returns(tminus4) 0117 (0035) 0111 (0030)Quarterly returns(tminus5) 0089 (0030) 0098 (0027)Quarterly returns(tminus6) 0054 (0033) 0066 (0030)Quarterly returns(tminus7) 0083 (0027) 0103 (0025)Quarterly returns(tminus8) 0044 (0022) 0060 (0022)Quarterly returns(tminus9) 0012 (0029) 0034 (0029)Quarterly returns(tminus10) minus0010 (0030) 0019 (0029)Quarterly returns(tminus11) 0025 (0027) 0056 (0026)Quarterly flow(tminus1) 0158 (0008)Log(Age + 1) minus0004 (0005)Log(AUM(tminus1)+1) minus0024 (0001)Volatility minus0141 (0037)Management fees 0211 (0386)Incentive fees 0005 (0053)HWM indicator 0020 (0004)Leveraged indicator 0000 (0004)Open-to-public indicator minus0002 (0005)On-shore indicator minus0032 (0005)Log(Total redemption period +1) 0034 (0007)Log(Lock-up period +1) minus0005 (0003)Log(Subscription period +1) minus0023 (0015)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0162 0194F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1940 (0000) 2077 (0000)
(Continued)
884 The Journal of Finance Rcopy
Table IImdashContinued
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel B Age interactions
Quarterly returns(tminus1)timesLog(Age in quarters+1) minus0319 (0035) minus0285 (0034)Quarterly returns(tminus2)timesLog(Age in quarters+1) minus0171 (0027) minus0125 (0025)Quarterly returns(tminus3)timesLog(Age in quarters+1) minus0051 (0036) minus0033 (0036)Quarterly returns(tminus4)timesLog(Age in quarters+1) minus0003 (0032) 0001 (0032)Quarterly returns(tminus5)timesLog(Age in quarters+1) minus0041 (0030) minus0042 (0029)Quarterly returns(tminus6)timesLog(Age in quarters+1) 0038 (0033) 0040 (0031)Quarterly returns(tminus7)timesLog(Age in quarters+1) 0013 (0033) 0007 (0032)Quarterly returns(tminus8)timesLog(Age in quarters+1) 0033 (0033) 0034 (0033)Quarterly returns(tminus9)timesLog(Age in quarters+1) 0003 (0040) minus0004 (0040)Quarterly returns(tminus10)timesLog(Age in quarters+1) 0132 (0038) 0133 (0035)Quarterly returns(tminus11)timesLog(Age in quarters+1) 0018 (0040) 0007 (0041)Quarterly returns(tminus1) 1325 (0117) 1233 (0113)Quarterly returns(tminus2) 0773 (0095) 0602 (0087)Quarterly returns(tminus3) 0354 (0114) 0287 (0110)Quarterly returns(tminus4) 0124 (0112) 0110 (0106)Quarterly returns(tminus5) 0200 (0100) 0215 (0092)Quarterly returns(tminus6) minus0067 (0112) minus0056 (0104)Quarterly returns(tminus7) 0033 (0106) 0079 (0099)Quarterly returns(tminus8) minus0066 (0107) minus0046 (0107)Quarterly returns(tminus9) 0004 (0128) 0053 (0127)Quarterly returns(tminus10) minus0416 (0122) minus0383 (0114)Quarterly returns(tminus11) minus0038 (0126) 0037 (0130)Log(Age + 1) minus0026 (0007) minus0000 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0168 0198F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1788 (0000) 1570 (0000)All age interaction terms 1214 (0000) 1084 (0000)
Panel C Scalability interactions
Quarterly returns(tminus1)timesConstrained minus0215 (0048) minus0215 (0047)Quarterly returns(tminus2)timesConstrained minus0096 (0043) minus0078 (0039)Quarterly returns(tminus3)timesConstrained minus0102 (0038) minus0107 (0037)Quarterly returns(tminus4)timesConstrained 0013 (0034) 0013 (0036)Quarterly returns(tminus5)timesConstrained minus0009 (0032) minus0022 (0032)Quarterly returns(tminus6)timesConstrained 0026 (0034) 0013 (0034)Quarterly returns(tminus7)timesConstrained minus0044 (0036) minus0057 (0036)Quarterly returns(tminus8)timesConstrained 0032 (0037) 0033 (0036)Quarterly returns(tminus9)timesConstrained minus0085 (0030) minus0090 (0031)Quarterly returns(tminus10)timesConstrained minus0023 (0036) minus0007 (0036)Quarterly returns(tminus11)timesConstrained minus0007 (0037) minus0004 (0038)
(Continued)
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 879
Our sample period is from January 1995 to December 2010 We focus onthe post-1994 period because TASS started reporting information on ldquodefunctrdquofunds only after 19945 We exclude managed futuresCTAs and funds-of-fundswhich have different incentive fees and are likely to have different inflow-performance relations from typical individual hedge funds We also excludeclosed-end hedge funds as subscriptions in these funds are only possible duringthe initial issuing period so future flows are not possible This initial filterleaves us with 4939 open-end hedge funds6
We drop funds for which TASS does not contain information on organizationalcharacteristics such as management fees incentive fees and HWM provisionsIn addition we consider only funds with an uninterrupted series of net assetvalues and returns so that we can calculate inflows Further we restrict oursample to funds with at least 12 consecutive monthly returns available duringthe sample period If a fund stops reporting returns and then resumes at afuture date we use only the first sequence of uninterrupted data Finally weexclude funds with an incentive fee of zero because there can be no direct payfor performance for these funds
We conduct the analysis using quarterly data because we wish to include onlylagged (not contemporaneous) returns in the flow-performance specificationsand to have a relatively short gap between returns and flows7 To construct aquarterly sample we drop all fund-calendar quarters that have return infor-mation only for a fraction of the period We also require that a sample fundhave a subscription period less than or equal to three months so that quarterlyinflows are not restricted This sample construction process leaves us with asample of 2998 funds (50333 fund-quarter observations)
Table I presents descriptive statistics Time-varying variables such as Quar-terly flow and Quarterly returns are measured at the fund-quarter level andother contractual characteristics such as Management fees and Incentive feesare measured at the fund level8 All time-varying variables except fund age(Age) are winsorized at the 1st and 99th percentiles to minimize the effect ofoutliers
5 Defunct funds include funds that are liquidated merged or restructured as well as those thatstopped reporting returns to TASS (Fung et al (2008))
6 Some funds are closed to new investors but unfortunately we do not know whether a particularfund is taking new money at any point in time so we cannot exclude funds on the basis of thispolicy Including closed funds causes us to understate the flow-performance relation for funds thatare not closed
7 Prior versions of this paper presented estimates using annual data with contemporaneousannual returns included in the flow-performance specifications (Lim Sensoy and Weisbach (2013))The estimates of indirect incentives using annual data are very close to those reported here
8 TASS provides information on fundsrsquo organizational characteristics as of the last available dateof fund data Like most previous studies we also assume that these organizational characteristicsdo not change throughout the life of the fund Agarwal Daniel and Naik (2009) argue that fundsrsquoorganizational characteristics are unlikely to change much over time based on their discussionswith practitioners which suggests that if a manager wants to impose new contractual terms it iseasier for him to start a new fund with different terms than to go through the legal complicationsof changing an existing contract
880 The Journal of Finance Rcopy
Mean Quarterly flow is 80 with a median of 03 so the distributionis highly skewed Mean Quarterly returns is 26 the median is 21 Aver-age fund size (AUM) is $2108 million The remaining variables reflect time-invariant contractual features Summary statistics on these characteristics arevery close to those reported in prior studies (eg Agarwal Daniel and Naik(2009) Baquero and Verbeek (2009) Aragon and Nanda (2012)) The variableManagement fees is the annual percentage of fund assets (AUM) received bythe manager as compensation and has a sample mean (median) of 15 (15)The variable Incentive fees is the percentage of profits above the HWM receivedby the manager as compensation and has a sample mean (median) of 193(20) Over two-thirds of our sample funds 694 have a HWM provision685 report that they use leverage 191 are open to public investors andabout a quarter are on-shore funds
The fee data consist of the fees that are currently publicly quoted by thefunds These data could potentially misrepresent the true fees relevant forour calculations for three reasons First funds sometimes provide fee reduc-tions to particular strategic investors that are not reflected in the database(Ramadorai and Streatfield (2011)) Although we cannot investigate this pos-sibility directly it is unlikely to have a major impact on our conclusions Forexample if the true incentive fee (management fee) averaged over all investorswere 10 lower than what is reported in the database our estimates of indirectincentives using the base GIR model would be overstated by about 1 (5) Asecond issue is that fees can change over time Agarwal and Ray (2012) andDueskar et al (2012) find that fee changes are infrequent and tend to reflectpast performance when they do occur so that fee increases follow good perfor-mance and fee decreases follow poor performance This effect is magnified inindirect incentives because good performance today leads not only to inflowsbut also to higher proportional fees on those inflows Third it is possible thatthere are fee changes unobservable to researchers To the extent that observ-able and unobservable fee changes are positively correlated this possibilityleads us to understate indirect incentives
We consider three variables that reflect potential restrictions on the be-havior of flows First we use Total redemption period defined as the sum ofthe notice period and the redemption period where the notice period is thetime the investor has to give notice to the fund about an intention to with-draw money from the fund and the redemption period is the time that thefund takes to return the money after the notice period is over The Lockupperiod is the minimum time that an investor has to wait before withdrawinginvested money The Subscription period is the time delay between investingin a fund and actually purchasing fund shares The mean total redemption pe-riod lock-up period and subscription period are 109 097 and 036 quartersrespectively
Indirect Incentives of Hedge Fund Managers 881
III Estimating the Sensitivity of Fund Inflows to Performance
To understand the impact of performance on fund flows we employ aBayesian learning framework that presumes investors are continually eval-uating managers trying to assess their abilities (see Berk and Green (2004)Chung et al (2012)) A fundrsquos performance provides information about the man-agerrsquos ability so an observation of performance will lead investors to updatetheir assessment of his ability and allocate more capital to a fund if their es-timate of the managerrsquos ability increases The magnitude of the updates andhence the sensitivity of inflows to performance should depend on the infor-mativeness of the signal relative to the precision of the prior estimate of thefund managerrsquos ability The sensitivity of inflows to performance should alsodepend on the extent to which ability can be scaled to replicate a fundrsquos returndistribution on new capital
Measuring the indirect incentives of hedge fund managers requires an es-timate of the relation between fund performance and future inflows A longliterature beginning with Ippolito (1992) estimates this relation to be rela-tively strong in the mutual fund industry9 It is also positive in the privateequity industry (see eg Kaplan and Schoar (2005)) However the results forhedge funds are less clear Goetzmann Ingersoll and Ross (2003) report a neg-ative and concave relation whereas other studies including Agarwal Danieland Naik (2003) Fung et al (2008) Baquero and Verbeek (2009) and Dinget al (2009) find a positive one
A Empirical Specification
We estimate the following specification
Flowit = β0 +11sumj=1
β0+ j Returnitminus j + γ Xtminus1 + λY + Fixed effects + εit (1)
where Flowit represents flows for fund i in quarter t10
Following the literature on flows to mutual funds (eg Sirri and Tufano(1998) Chevalier and Ellison (1999)) or hedge funds (eg Fung et al (2008)Agarwal Daniel and Naik (2009)) we compute quarterly flows of capital intoa fund as follows
Flowit = AUMit minus AUMitminus1(1 + Returnit
)AUMitminus1
(2)
where AUMit and AUMitminus1 are the AUM of fund i at the end of quarters t andtminus1 respectively and Returnit is the net-of-fee return for fund i in quarter t
9 See Brown Halow and Starks (1996) Sirri and Tufano (1998) Chevalier and Ellison (1999)Barclay Pearson and Weisbach (1998) Del Guercio and Tkac (2002) Bollen (2007) Huang Weiand Yan (2007) and Sensoy (2009)
10 We restrict our estimates to quarterly specifications but prior versions of this paper alsoinclude annual and monthly specifications with similar results to those reported below
882 The Journal of Finance Rcopy
This definition expresses flows as a fraction of beginning-of-period (end ofprior period) AUM which is a natural benchmark from the perspective of afund manager assessing his or her incentives going forward For instance theoption deltas that comprise direct incentives are defined in terms of beginning-of-period AUM An alternative approach to computing fund flows is to scalethe denominator of equation (2) by (1 + Returnit) which expresses flows as afraction of what end-of-period AUM would have been in the absence of flowsAlthough this alternative definition is less intuitive for our purpose it leads tosimilar estimates of indirect incentives
The vector X in equation (1) consists of time-varying fund characteristicsthat include lagged flows the natural logarithm of AUM for fund i at tminus1 thenatural logarithm of one plus fund irsquos age in quarters at tminus1 and fund volatil-ity over the previous 12 months11 The vector Y includes time-invariant fundcharacteristics that include Management fees Incentive fees Total redemptionperiod Lockup period Subscription period and a set of indicator variables thatequal one if the fund has an HWM provision uses leverage is open to publicinvestors and is an on-shore fund All specifications include fixed effects forthe nine styles listed in Table I interacted with calendar quarter fixed effectsThese time-by-style effects capture all shocks observed or unobserved that arecommon to funds of a given style in a given quarter including the returns topeer funds and inflows to other funds of the same style12 Reported standarderrors are robust to heteroskedasticity and account for double clustering byfund and time period13
B Estimates of the Flow-Performance Relation
We present estimates of the flow-performance relation in Panel A ofTable II In Column (1) we include returns in the 11 quarters prior to thecurrent quarter In Column (2) we add a number of fund-level controls In eachspecification the coefficients on returns are positive and statistically signif-icant in most cases and decline sharply over time so the coefficient on themost recent quarterrsquos return is the largest If we sum the coefficients on the 11prior quarters the sum in Column (1) is 144 and in Column (2) is 150 Thesecoefficients imply that a one-standard-deviation increase in returns (9) willlead to about a 13 increase in fund size
11 For young funds that have for example only one yearrsquos worth of return history we cannotcompute lagged returns and flows In such cases we ldquodummy outrdquo missing lagged variables toretain observations To do so we set missing values of lagged flows and returns to zero and includean indicator for missing values
12 Time-by-style fixed effects perform the same adjustment as a factor model regression underthe assumption that the factor loadings are the same for all funds of a given style within a giventime period
13 We focus on linear specifications Although there is some evidence of nonlinear effects in thedata (using splines quadratics etc) they are small in magnitude and have little impact on theestimates obtained with linear specifications
Indirect Incentives of Hedge Fund Managers 883
Table IIEstimates of Flow-Performance Sensitivity
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) using quarterly data The dependent variable is Quarterly flow as defined byequation (2) See Table I for definitions of all independent variables The sample period is from 1995to 2010 We only employ the funds with a subscription period less than or equal to three months toavoid quarterly inflows being restricted by subscription periods Equation (1) is augmented in PanelB to include interaction terms of the log of the fundrsquos age plus one with prior performance and inPanel C to include interaction terms of Constrained with performance variables All specificationsin Panel A and B include style fixed effects quarter fixed effects and quarter-by-style fixed effectsOnly quarter fixed effects are included in Panel C due to multicollinearity Standard errors aredouble clustered by fund and year and indicate statistical significance at the 1 5and 10 level respectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel A Base-case
Quarterly returns(tminus1) 0501 (0045) 0496 (0043)Quarterly returns(tminus2) 0314 (0039) 0263 (0036)Quarterly returns(tminus3) 0209 (0034) 0190 (0028)Quarterly returns(tminus4) 0117 (0035) 0111 (0030)Quarterly returns(tminus5) 0089 (0030) 0098 (0027)Quarterly returns(tminus6) 0054 (0033) 0066 (0030)Quarterly returns(tminus7) 0083 (0027) 0103 (0025)Quarterly returns(tminus8) 0044 (0022) 0060 (0022)Quarterly returns(tminus9) 0012 (0029) 0034 (0029)Quarterly returns(tminus10) minus0010 (0030) 0019 (0029)Quarterly returns(tminus11) 0025 (0027) 0056 (0026)Quarterly flow(tminus1) 0158 (0008)Log(Age + 1) minus0004 (0005)Log(AUM(tminus1)+1) minus0024 (0001)Volatility minus0141 (0037)Management fees 0211 (0386)Incentive fees 0005 (0053)HWM indicator 0020 (0004)Leveraged indicator 0000 (0004)Open-to-public indicator minus0002 (0005)On-shore indicator minus0032 (0005)Log(Total redemption period +1) 0034 (0007)Log(Lock-up period +1) minus0005 (0003)Log(Subscription period +1) minus0023 (0015)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0162 0194F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1940 (0000) 2077 (0000)
(Continued)
884 The Journal of Finance Rcopy
Table IImdashContinued
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel B Age interactions
Quarterly returns(tminus1)timesLog(Age in quarters+1) minus0319 (0035) minus0285 (0034)Quarterly returns(tminus2)timesLog(Age in quarters+1) minus0171 (0027) minus0125 (0025)Quarterly returns(tminus3)timesLog(Age in quarters+1) minus0051 (0036) minus0033 (0036)Quarterly returns(tminus4)timesLog(Age in quarters+1) minus0003 (0032) 0001 (0032)Quarterly returns(tminus5)timesLog(Age in quarters+1) minus0041 (0030) minus0042 (0029)Quarterly returns(tminus6)timesLog(Age in quarters+1) 0038 (0033) 0040 (0031)Quarterly returns(tminus7)timesLog(Age in quarters+1) 0013 (0033) 0007 (0032)Quarterly returns(tminus8)timesLog(Age in quarters+1) 0033 (0033) 0034 (0033)Quarterly returns(tminus9)timesLog(Age in quarters+1) 0003 (0040) minus0004 (0040)Quarterly returns(tminus10)timesLog(Age in quarters+1) 0132 (0038) 0133 (0035)Quarterly returns(tminus11)timesLog(Age in quarters+1) 0018 (0040) 0007 (0041)Quarterly returns(tminus1) 1325 (0117) 1233 (0113)Quarterly returns(tminus2) 0773 (0095) 0602 (0087)Quarterly returns(tminus3) 0354 (0114) 0287 (0110)Quarterly returns(tminus4) 0124 (0112) 0110 (0106)Quarterly returns(tminus5) 0200 (0100) 0215 (0092)Quarterly returns(tminus6) minus0067 (0112) minus0056 (0104)Quarterly returns(tminus7) 0033 (0106) 0079 (0099)Quarterly returns(tminus8) minus0066 (0107) minus0046 (0107)Quarterly returns(tminus9) 0004 (0128) 0053 (0127)Quarterly returns(tminus10) minus0416 (0122) minus0383 (0114)Quarterly returns(tminus11) minus0038 (0126) 0037 (0130)Log(Age + 1) minus0026 (0007) minus0000 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0168 0198F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1788 (0000) 1570 (0000)All age interaction terms 1214 (0000) 1084 (0000)
Panel C Scalability interactions
Quarterly returns(tminus1)timesConstrained minus0215 (0048) minus0215 (0047)Quarterly returns(tminus2)timesConstrained minus0096 (0043) minus0078 (0039)Quarterly returns(tminus3)timesConstrained minus0102 (0038) minus0107 (0037)Quarterly returns(tminus4)timesConstrained 0013 (0034) 0013 (0036)Quarterly returns(tminus5)timesConstrained minus0009 (0032) minus0022 (0032)Quarterly returns(tminus6)timesConstrained 0026 (0034) 0013 (0034)Quarterly returns(tminus7)timesConstrained minus0044 (0036) minus0057 (0036)Quarterly returns(tminus8)timesConstrained 0032 (0037) 0033 (0036)Quarterly returns(tminus9)timesConstrained minus0085 (0030) minus0090 (0031)Quarterly returns(tminus10)timesConstrained minus0023 (0036) minus0007 (0036)Quarterly returns(tminus11)timesConstrained minus0007 (0037) minus0004 (0038)
(Continued)
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
880 The Journal of Finance Rcopy
Mean Quarterly flow is 80 with a median of 03 so the distributionis highly skewed Mean Quarterly returns is 26 the median is 21 Aver-age fund size (AUM) is $2108 million The remaining variables reflect time-invariant contractual features Summary statistics on these characteristics arevery close to those reported in prior studies (eg Agarwal Daniel and Naik(2009) Baquero and Verbeek (2009) Aragon and Nanda (2012)) The variableManagement fees is the annual percentage of fund assets (AUM) received bythe manager as compensation and has a sample mean (median) of 15 (15)The variable Incentive fees is the percentage of profits above the HWM receivedby the manager as compensation and has a sample mean (median) of 193(20) Over two-thirds of our sample funds 694 have a HWM provision685 report that they use leverage 191 are open to public investors andabout a quarter are on-shore funds
The fee data consist of the fees that are currently publicly quoted by thefunds These data could potentially misrepresent the true fees relevant forour calculations for three reasons First funds sometimes provide fee reduc-tions to particular strategic investors that are not reflected in the database(Ramadorai and Streatfield (2011)) Although we cannot investigate this pos-sibility directly it is unlikely to have a major impact on our conclusions Forexample if the true incentive fee (management fee) averaged over all investorswere 10 lower than what is reported in the database our estimates of indirectincentives using the base GIR model would be overstated by about 1 (5) Asecond issue is that fees can change over time Agarwal and Ray (2012) andDueskar et al (2012) find that fee changes are infrequent and tend to reflectpast performance when they do occur so that fee increases follow good perfor-mance and fee decreases follow poor performance This effect is magnified inindirect incentives because good performance today leads not only to inflowsbut also to higher proportional fees on those inflows Third it is possible thatthere are fee changes unobservable to researchers To the extent that observ-able and unobservable fee changes are positively correlated this possibilityleads us to understate indirect incentives
We consider three variables that reflect potential restrictions on the be-havior of flows First we use Total redemption period defined as the sum ofthe notice period and the redemption period where the notice period is thetime the investor has to give notice to the fund about an intention to with-draw money from the fund and the redemption period is the time that thefund takes to return the money after the notice period is over The Lockupperiod is the minimum time that an investor has to wait before withdrawinginvested money The Subscription period is the time delay between investingin a fund and actually purchasing fund shares The mean total redemption pe-riod lock-up period and subscription period are 109 097 and 036 quartersrespectively
Indirect Incentives of Hedge Fund Managers 881
III Estimating the Sensitivity of Fund Inflows to Performance
To understand the impact of performance on fund flows we employ aBayesian learning framework that presumes investors are continually eval-uating managers trying to assess their abilities (see Berk and Green (2004)Chung et al (2012)) A fundrsquos performance provides information about the man-agerrsquos ability so an observation of performance will lead investors to updatetheir assessment of his ability and allocate more capital to a fund if their es-timate of the managerrsquos ability increases The magnitude of the updates andhence the sensitivity of inflows to performance should depend on the infor-mativeness of the signal relative to the precision of the prior estimate of thefund managerrsquos ability The sensitivity of inflows to performance should alsodepend on the extent to which ability can be scaled to replicate a fundrsquos returndistribution on new capital
Measuring the indirect incentives of hedge fund managers requires an es-timate of the relation between fund performance and future inflows A longliterature beginning with Ippolito (1992) estimates this relation to be rela-tively strong in the mutual fund industry9 It is also positive in the privateequity industry (see eg Kaplan and Schoar (2005)) However the results forhedge funds are less clear Goetzmann Ingersoll and Ross (2003) report a neg-ative and concave relation whereas other studies including Agarwal Danieland Naik (2003) Fung et al (2008) Baquero and Verbeek (2009) and Dinget al (2009) find a positive one
A Empirical Specification
We estimate the following specification
Flowit = β0 +11sumj=1
β0+ j Returnitminus j + γ Xtminus1 + λY + Fixed effects + εit (1)
where Flowit represents flows for fund i in quarter t10
Following the literature on flows to mutual funds (eg Sirri and Tufano(1998) Chevalier and Ellison (1999)) or hedge funds (eg Fung et al (2008)Agarwal Daniel and Naik (2009)) we compute quarterly flows of capital intoa fund as follows
Flowit = AUMit minus AUMitminus1(1 + Returnit
)AUMitminus1
(2)
where AUMit and AUMitminus1 are the AUM of fund i at the end of quarters t andtminus1 respectively and Returnit is the net-of-fee return for fund i in quarter t
9 See Brown Halow and Starks (1996) Sirri and Tufano (1998) Chevalier and Ellison (1999)Barclay Pearson and Weisbach (1998) Del Guercio and Tkac (2002) Bollen (2007) Huang Weiand Yan (2007) and Sensoy (2009)
10 We restrict our estimates to quarterly specifications but prior versions of this paper alsoinclude annual and monthly specifications with similar results to those reported below
882 The Journal of Finance Rcopy
This definition expresses flows as a fraction of beginning-of-period (end ofprior period) AUM which is a natural benchmark from the perspective of afund manager assessing his or her incentives going forward For instance theoption deltas that comprise direct incentives are defined in terms of beginning-of-period AUM An alternative approach to computing fund flows is to scalethe denominator of equation (2) by (1 + Returnit) which expresses flows as afraction of what end-of-period AUM would have been in the absence of flowsAlthough this alternative definition is less intuitive for our purpose it leads tosimilar estimates of indirect incentives
The vector X in equation (1) consists of time-varying fund characteristicsthat include lagged flows the natural logarithm of AUM for fund i at tminus1 thenatural logarithm of one plus fund irsquos age in quarters at tminus1 and fund volatil-ity over the previous 12 months11 The vector Y includes time-invariant fundcharacteristics that include Management fees Incentive fees Total redemptionperiod Lockup period Subscription period and a set of indicator variables thatequal one if the fund has an HWM provision uses leverage is open to publicinvestors and is an on-shore fund All specifications include fixed effects forthe nine styles listed in Table I interacted with calendar quarter fixed effectsThese time-by-style effects capture all shocks observed or unobserved that arecommon to funds of a given style in a given quarter including the returns topeer funds and inflows to other funds of the same style12 Reported standarderrors are robust to heteroskedasticity and account for double clustering byfund and time period13
B Estimates of the Flow-Performance Relation
We present estimates of the flow-performance relation in Panel A ofTable II In Column (1) we include returns in the 11 quarters prior to thecurrent quarter In Column (2) we add a number of fund-level controls In eachspecification the coefficients on returns are positive and statistically signif-icant in most cases and decline sharply over time so the coefficient on themost recent quarterrsquos return is the largest If we sum the coefficients on the 11prior quarters the sum in Column (1) is 144 and in Column (2) is 150 Thesecoefficients imply that a one-standard-deviation increase in returns (9) willlead to about a 13 increase in fund size
11 For young funds that have for example only one yearrsquos worth of return history we cannotcompute lagged returns and flows In such cases we ldquodummy outrdquo missing lagged variables toretain observations To do so we set missing values of lagged flows and returns to zero and includean indicator for missing values
12 Time-by-style fixed effects perform the same adjustment as a factor model regression underthe assumption that the factor loadings are the same for all funds of a given style within a giventime period
13 We focus on linear specifications Although there is some evidence of nonlinear effects in thedata (using splines quadratics etc) they are small in magnitude and have little impact on theestimates obtained with linear specifications
Indirect Incentives of Hedge Fund Managers 883
Table IIEstimates of Flow-Performance Sensitivity
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) using quarterly data The dependent variable is Quarterly flow as defined byequation (2) See Table I for definitions of all independent variables The sample period is from 1995to 2010 We only employ the funds with a subscription period less than or equal to three months toavoid quarterly inflows being restricted by subscription periods Equation (1) is augmented in PanelB to include interaction terms of the log of the fundrsquos age plus one with prior performance and inPanel C to include interaction terms of Constrained with performance variables All specificationsin Panel A and B include style fixed effects quarter fixed effects and quarter-by-style fixed effectsOnly quarter fixed effects are included in Panel C due to multicollinearity Standard errors aredouble clustered by fund and year and indicate statistical significance at the 1 5and 10 level respectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel A Base-case
Quarterly returns(tminus1) 0501 (0045) 0496 (0043)Quarterly returns(tminus2) 0314 (0039) 0263 (0036)Quarterly returns(tminus3) 0209 (0034) 0190 (0028)Quarterly returns(tminus4) 0117 (0035) 0111 (0030)Quarterly returns(tminus5) 0089 (0030) 0098 (0027)Quarterly returns(tminus6) 0054 (0033) 0066 (0030)Quarterly returns(tminus7) 0083 (0027) 0103 (0025)Quarterly returns(tminus8) 0044 (0022) 0060 (0022)Quarterly returns(tminus9) 0012 (0029) 0034 (0029)Quarterly returns(tminus10) minus0010 (0030) 0019 (0029)Quarterly returns(tminus11) 0025 (0027) 0056 (0026)Quarterly flow(tminus1) 0158 (0008)Log(Age + 1) minus0004 (0005)Log(AUM(tminus1)+1) minus0024 (0001)Volatility minus0141 (0037)Management fees 0211 (0386)Incentive fees 0005 (0053)HWM indicator 0020 (0004)Leveraged indicator 0000 (0004)Open-to-public indicator minus0002 (0005)On-shore indicator minus0032 (0005)Log(Total redemption period +1) 0034 (0007)Log(Lock-up period +1) minus0005 (0003)Log(Subscription period +1) minus0023 (0015)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0162 0194F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1940 (0000) 2077 (0000)
(Continued)
884 The Journal of Finance Rcopy
Table IImdashContinued
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel B Age interactions
Quarterly returns(tminus1)timesLog(Age in quarters+1) minus0319 (0035) minus0285 (0034)Quarterly returns(tminus2)timesLog(Age in quarters+1) minus0171 (0027) minus0125 (0025)Quarterly returns(tminus3)timesLog(Age in quarters+1) minus0051 (0036) minus0033 (0036)Quarterly returns(tminus4)timesLog(Age in quarters+1) minus0003 (0032) 0001 (0032)Quarterly returns(tminus5)timesLog(Age in quarters+1) minus0041 (0030) minus0042 (0029)Quarterly returns(tminus6)timesLog(Age in quarters+1) 0038 (0033) 0040 (0031)Quarterly returns(tminus7)timesLog(Age in quarters+1) 0013 (0033) 0007 (0032)Quarterly returns(tminus8)timesLog(Age in quarters+1) 0033 (0033) 0034 (0033)Quarterly returns(tminus9)timesLog(Age in quarters+1) 0003 (0040) minus0004 (0040)Quarterly returns(tminus10)timesLog(Age in quarters+1) 0132 (0038) 0133 (0035)Quarterly returns(tminus11)timesLog(Age in quarters+1) 0018 (0040) 0007 (0041)Quarterly returns(tminus1) 1325 (0117) 1233 (0113)Quarterly returns(tminus2) 0773 (0095) 0602 (0087)Quarterly returns(tminus3) 0354 (0114) 0287 (0110)Quarterly returns(tminus4) 0124 (0112) 0110 (0106)Quarterly returns(tminus5) 0200 (0100) 0215 (0092)Quarterly returns(tminus6) minus0067 (0112) minus0056 (0104)Quarterly returns(tminus7) 0033 (0106) 0079 (0099)Quarterly returns(tminus8) minus0066 (0107) minus0046 (0107)Quarterly returns(tminus9) 0004 (0128) 0053 (0127)Quarterly returns(tminus10) minus0416 (0122) minus0383 (0114)Quarterly returns(tminus11) minus0038 (0126) 0037 (0130)Log(Age + 1) minus0026 (0007) minus0000 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0168 0198F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1788 (0000) 1570 (0000)All age interaction terms 1214 (0000) 1084 (0000)
Panel C Scalability interactions
Quarterly returns(tminus1)timesConstrained minus0215 (0048) minus0215 (0047)Quarterly returns(tminus2)timesConstrained minus0096 (0043) minus0078 (0039)Quarterly returns(tminus3)timesConstrained minus0102 (0038) minus0107 (0037)Quarterly returns(tminus4)timesConstrained 0013 (0034) 0013 (0036)Quarterly returns(tminus5)timesConstrained minus0009 (0032) minus0022 (0032)Quarterly returns(tminus6)timesConstrained 0026 (0034) 0013 (0034)Quarterly returns(tminus7)timesConstrained minus0044 (0036) minus0057 (0036)Quarterly returns(tminus8)timesConstrained 0032 (0037) 0033 (0036)Quarterly returns(tminus9)timesConstrained minus0085 (0030) minus0090 (0031)Quarterly returns(tminus10)timesConstrained minus0023 (0036) minus0007 (0036)Quarterly returns(tminus11)timesConstrained minus0007 (0037) minus0004 (0038)
(Continued)
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 881
III Estimating the Sensitivity of Fund Inflows to Performance
To understand the impact of performance on fund flows we employ aBayesian learning framework that presumes investors are continually eval-uating managers trying to assess their abilities (see Berk and Green (2004)Chung et al (2012)) A fundrsquos performance provides information about the man-agerrsquos ability so an observation of performance will lead investors to updatetheir assessment of his ability and allocate more capital to a fund if their es-timate of the managerrsquos ability increases The magnitude of the updates andhence the sensitivity of inflows to performance should depend on the infor-mativeness of the signal relative to the precision of the prior estimate of thefund managerrsquos ability The sensitivity of inflows to performance should alsodepend on the extent to which ability can be scaled to replicate a fundrsquos returndistribution on new capital
Measuring the indirect incentives of hedge fund managers requires an es-timate of the relation between fund performance and future inflows A longliterature beginning with Ippolito (1992) estimates this relation to be rela-tively strong in the mutual fund industry9 It is also positive in the privateequity industry (see eg Kaplan and Schoar (2005)) However the results forhedge funds are less clear Goetzmann Ingersoll and Ross (2003) report a neg-ative and concave relation whereas other studies including Agarwal Danieland Naik (2003) Fung et al (2008) Baquero and Verbeek (2009) and Dinget al (2009) find a positive one
A Empirical Specification
We estimate the following specification
Flowit = β0 +11sumj=1
β0+ j Returnitminus j + γ Xtminus1 + λY + Fixed effects + εit (1)
where Flowit represents flows for fund i in quarter t10
Following the literature on flows to mutual funds (eg Sirri and Tufano(1998) Chevalier and Ellison (1999)) or hedge funds (eg Fung et al (2008)Agarwal Daniel and Naik (2009)) we compute quarterly flows of capital intoa fund as follows
Flowit = AUMit minus AUMitminus1(1 + Returnit
)AUMitminus1
(2)
where AUMit and AUMitminus1 are the AUM of fund i at the end of quarters t andtminus1 respectively and Returnit is the net-of-fee return for fund i in quarter t
9 See Brown Halow and Starks (1996) Sirri and Tufano (1998) Chevalier and Ellison (1999)Barclay Pearson and Weisbach (1998) Del Guercio and Tkac (2002) Bollen (2007) Huang Weiand Yan (2007) and Sensoy (2009)
10 We restrict our estimates to quarterly specifications but prior versions of this paper alsoinclude annual and monthly specifications with similar results to those reported below
882 The Journal of Finance Rcopy
This definition expresses flows as a fraction of beginning-of-period (end ofprior period) AUM which is a natural benchmark from the perspective of afund manager assessing his or her incentives going forward For instance theoption deltas that comprise direct incentives are defined in terms of beginning-of-period AUM An alternative approach to computing fund flows is to scalethe denominator of equation (2) by (1 + Returnit) which expresses flows as afraction of what end-of-period AUM would have been in the absence of flowsAlthough this alternative definition is less intuitive for our purpose it leads tosimilar estimates of indirect incentives
The vector X in equation (1) consists of time-varying fund characteristicsthat include lagged flows the natural logarithm of AUM for fund i at tminus1 thenatural logarithm of one plus fund irsquos age in quarters at tminus1 and fund volatil-ity over the previous 12 months11 The vector Y includes time-invariant fundcharacteristics that include Management fees Incentive fees Total redemptionperiod Lockup period Subscription period and a set of indicator variables thatequal one if the fund has an HWM provision uses leverage is open to publicinvestors and is an on-shore fund All specifications include fixed effects forthe nine styles listed in Table I interacted with calendar quarter fixed effectsThese time-by-style effects capture all shocks observed or unobserved that arecommon to funds of a given style in a given quarter including the returns topeer funds and inflows to other funds of the same style12 Reported standarderrors are robust to heteroskedasticity and account for double clustering byfund and time period13
B Estimates of the Flow-Performance Relation
We present estimates of the flow-performance relation in Panel A ofTable II In Column (1) we include returns in the 11 quarters prior to thecurrent quarter In Column (2) we add a number of fund-level controls In eachspecification the coefficients on returns are positive and statistically signif-icant in most cases and decline sharply over time so the coefficient on themost recent quarterrsquos return is the largest If we sum the coefficients on the 11prior quarters the sum in Column (1) is 144 and in Column (2) is 150 Thesecoefficients imply that a one-standard-deviation increase in returns (9) willlead to about a 13 increase in fund size
11 For young funds that have for example only one yearrsquos worth of return history we cannotcompute lagged returns and flows In such cases we ldquodummy outrdquo missing lagged variables toretain observations To do so we set missing values of lagged flows and returns to zero and includean indicator for missing values
12 Time-by-style fixed effects perform the same adjustment as a factor model regression underthe assumption that the factor loadings are the same for all funds of a given style within a giventime period
13 We focus on linear specifications Although there is some evidence of nonlinear effects in thedata (using splines quadratics etc) they are small in magnitude and have little impact on theestimates obtained with linear specifications
Indirect Incentives of Hedge Fund Managers 883
Table IIEstimates of Flow-Performance Sensitivity
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) using quarterly data The dependent variable is Quarterly flow as defined byequation (2) See Table I for definitions of all independent variables The sample period is from 1995to 2010 We only employ the funds with a subscription period less than or equal to three months toavoid quarterly inflows being restricted by subscription periods Equation (1) is augmented in PanelB to include interaction terms of the log of the fundrsquos age plus one with prior performance and inPanel C to include interaction terms of Constrained with performance variables All specificationsin Panel A and B include style fixed effects quarter fixed effects and quarter-by-style fixed effectsOnly quarter fixed effects are included in Panel C due to multicollinearity Standard errors aredouble clustered by fund and year and indicate statistical significance at the 1 5and 10 level respectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel A Base-case
Quarterly returns(tminus1) 0501 (0045) 0496 (0043)Quarterly returns(tminus2) 0314 (0039) 0263 (0036)Quarterly returns(tminus3) 0209 (0034) 0190 (0028)Quarterly returns(tminus4) 0117 (0035) 0111 (0030)Quarterly returns(tminus5) 0089 (0030) 0098 (0027)Quarterly returns(tminus6) 0054 (0033) 0066 (0030)Quarterly returns(tminus7) 0083 (0027) 0103 (0025)Quarterly returns(tminus8) 0044 (0022) 0060 (0022)Quarterly returns(tminus9) 0012 (0029) 0034 (0029)Quarterly returns(tminus10) minus0010 (0030) 0019 (0029)Quarterly returns(tminus11) 0025 (0027) 0056 (0026)Quarterly flow(tminus1) 0158 (0008)Log(Age + 1) minus0004 (0005)Log(AUM(tminus1)+1) minus0024 (0001)Volatility minus0141 (0037)Management fees 0211 (0386)Incentive fees 0005 (0053)HWM indicator 0020 (0004)Leveraged indicator 0000 (0004)Open-to-public indicator minus0002 (0005)On-shore indicator minus0032 (0005)Log(Total redemption period +1) 0034 (0007)Log(Lock-up period +1) minus0005 (0003)Log(Subscription period +1) minus0023 (0015)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0162 0194F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1940 (0000) 2077 (0000)
(Continued)
884 The Journal of Finance Rcopy
Table IImdashContinued
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel B Age interactions
Quarterly returns(tminus1)timesLog(Age in quarters+1) minus0319 (0035) minus0285 (0034)Quarterly returns(tminus2)timesLog(Age in quarters+1) minus0171 (0027) minus0125 (0025)Quarterly returns(tminus3)timesLog(Age in quarters+1) minus0051 (0036) minus0033 (0036)Quarterly returns(tminus4)timesLog(Age in quarters+1) minus0003 (0032) 0001 (0032)Quarterly returns(tminus5)timesLog(Age in quarters+1) minus0041 (0030) minus0042 (0029)Quarterly returns(tminus6)timesLog(Age in quarters+1) 0038 (0033) 0040 (0031)Quarterly returns(tminus7)timesLog(Age in quarters+1) 0013 (0033) 0007 (0032)Quarterly returns(tminus8)timesLog(Age in quarters+1) 0033 (0033) 0034 (0033)Quarterly returns(tminus9)timesLog(Age in quarters+1) 0003 (0040) minus0004 (0040)Quarterly returns(tminus10)timesLog(Age in quarters+1) 0132 (0038) 0133 (0035)Quarterly returns(tminus11)timesLog(Age in quarters+1) 0018 (0040) 0007 (0041)Quarterly returns(tminus1) 1325 (0117) 1233 (0113)Quarterly returns(tminus2) 0773 (0095) 0602 (0087)Quarterly returns(tminus3) 0354 (0114) 0287 (0110)Quarterly returns(tminus4) 0124 (0112) 0110 (0106)Quarterly returns(tminus5) 0200 (0100) 0215 (0092)Quarterly returns(tminus6) minus0067 (0112) minus0056 (0104)Quarterly returns(tminus7) 0033 (0106) 0079 (0099)Quarterly returns(tminus8) minus0066 (0107) minus0046 (0107)Quarterly returns(tminus9) 0004 (0128) 0053 (0127)Quarterly returns(tminus10) minus0416 (0122) minus0383 (0114)Quarterly returns(tminus11) minus0038 (0126) 0037 (0130)Log(Age + 1) minus0026 (0007) minus0000 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0168 0198F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1788 (0000) 1570 (0000)All age interaction terms 1214 (0000) 1084 (0000)
Panel C Scalability interactions
Quarterly returns(tminus1)timesConstrained minus0215 (0048) minus0215 (0047)Quarterly returns(tminus2)timesConstrained minus0096 (0043) minus0078 (0039)Quarterly returns(tminus3)timesConstrained minus0102 (0038) minus0107 (0037)Quarterly returns(tminus4)timesConstrained 0013 (0034) 0013 (0036)Quarterly returns(tminus5)timesConstrained minus0009 (0032) minus0022 (0032)Quarterly returns(tminus6)timesConstrained 0026 (0034) 0013 (0034)Quarterly returns(tminus7)timesConstrained minus0044 (0036) minus0057 (0036)Quarterly returns(tminus8)timesConstrained 0032 (0037) 0033 (0036)Quarterly returns(tminus9)timesConstrained minus0085 (0030) minus0090 (0031)Quarterly returns(tminus10)timesConstrained minus0023 (0036) minus0007 (0036)Quarterly returns(tminus11)timesConstrained minus0007 (0037) minus0004 (0038)
(Continued)
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
882 The Journal of Finance Rcopy
This definition expresses flows as a fraction of beginning-of-period (end ofprior period) AUM which is a natural benchmark from the perspective of afund manager assessing his or her incentives going forward For instance theoption deltas that comprise direct incentives are defined in terms of beginning-of-period AUM An alternative approach to computing fund flows is to scalethe denominator of equation (2) by (1 + Returnit) which expresses flows as afraction of what end-of-period AUM would have been in the absence of flowsAlthough this alternative definition is less intuitive for our purpose it leads tosimilar estimates of indirect incentives
The vector X in equation (1) consists of time-varying fund characteristicsthat include lagged flows the natural logarithm of AUM for fund i at tminus1 thenatural logarithm of one plus fund irsquos age in quarters at tminus1 and fund volatil-ity over the previous 12 months11 The vector Y includes time-invariant fundcharacteristics that include Management fees Incentive fees Total redemptionperiod Lockup period Subscription period and a set of indicator variables thatequal one if the fund has an HWM provision uses leverage is open to publicinvestors and is an on-shore fund All specifications include fixed effects forthe nine styles listed in Table I interacted with calendar quarter fixed effectsThese time-by-style effects capture all shocks observed or unobserved that arecommon to funds of a given style in a given quarter including the returns topeer funds and inflows to other funds of the same style12 Reported standarderrors are robust to heteroskedasticity and account for double clustering byfund and time period13
B Estimates of the Flow-Performance Relation
We present estimates of the flow-performance relation in Panel A ofTable II In Column (1) we include returns in the 11 quarters prior to thecurrent quarter In Column (2) we add a number of fund-level controls In eachspecification the coefficients on returns are positive and statistically signif-icant in most cases and decline sharply over time so the coefficient on themost recent quarterrsquos return is the largest If we sum the coefficients on the 11prior quarters the sum in Column (1) is 144 and in Column (2) is 150 Thesecoefficients imply that a one-standard-deviation increase in returns (9) willlead to about a 13 increase in fund size
11 For young funds that have for example only one yearrsquos worth of return history we cannotcompute lagged returns and flows In such cases we ldquodummy outrdquo missing lagged variables toretain observations To do so we set missing values of lagged flows and returns to zero and includean indicator for missing values
12 Time-by-style fixed effects perform the same adjustment as a factor model regression underthe assumption that the factor loadings are the same for all funds of a given style within a giventime period
13 We focus on linear specifications Although there is some evidence of nonlinear effects in thedata (using splines quadratics etc) they are small in magnitude and have little impact on theestimates obtained with linear specifications
Indirect Incentives of Hedge Fund Managers 883
Table IIEstimates of Flow-Performance Sensitivity
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) using quarterly data The dependent variable is Quarterly flow as defined byequation (2) See Table I for definitions of all independent variables The sample period is from 1995to 2010 We only employ the funds with a subscription period less than or equal to three months toavoid quarterly inflows being restricted by subscription periods Equation (1) is augmented in PanelB to include interaction terms of the log of the fundrsquos age plus one with prior performance and inPanel C to include interaction terms of Constrained with performance variables All specificationsin Panel A and B include style fixed effects quarter fixed effects and quarter-by-style fixed effectsOnly quarter fixed effects are included in Panel C due to multicollinearity Standard errors aredouble clustered by fund and year and indicate statistical significance at the 1 5and 10 level respectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel A Base-case
Quarterly returns(tminus1) 0501 (0045) 0496 (0043)Quarterly returns(tminus2) 0314 (0039) 0263 (0036)Quarterly returns(tminus3) 0209 (0034) 0190 (0028)Quarterly returns(tminus4) 0117 (0035) 0111 (0030)Quarterly returns(tminus5) 0089 (0030) 0098 (0027)Quarterly returns(tminus6) 0054 (0033) 0066 (0030)Quarterly returns(tminus7) 0083 (0027) 0103 (0025)Quarterly returns(tminus8) 0044 (0022) 0060 (0022)Quarterly returns(tminus9) 0012 (0029) 0034 (0029)Quarterly returns(tminus10) minus0010 (0030) 0019 (0029)Quarterly returns(tminus11) 0025 (0027) 0056 (0026)Quarterly flow(tminus1) 0158 (0008)Log(Age + 1) minus0004 (0005)Log(AUM(tminus1)+1) minus0024 (0001)Volatility minus0141 (0037)Management fees 0211 (0386)Incentive fees 0005 (0053)HWM indicator 0020 (0004)Leveraged indicator 0000 (0004)Open-to-public indicator minus0002 (0005)On-shore indicator minus0032 (0005)Log(Total redemption period +1) 0034 (0007)Log(Lock-up period +1) minus0005 (0003)Log(Subscription period +1) minus0023 (0015)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0162 0194F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1940 (0000) 2077 (0000)
(Continued)
884 The Journal of Finance Rcopy
Table IImdashContinued
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel B Age interactions
Quarterly returns(tminus1)timesLog(Age in quarters+1) minus0319 (0035) minus0285 (0034)Quarterly returns(tminus2)timesLog(Age in quarters+1) minus0171 (0027) minus0125 (0025)Quarterly returns(tminus3)timesLog(Age in quarters+1) minus0051 (0036) minus0033 (0036)Quarterly returns(tminus4)timesLog(Age in quarters+1) minus0003 (0032) 0001 (0032)Quarterly returns(tminus5)timesLog(Age in quarters+1) minus0041 (0030) minus0042 (0029)Quarterly returns(tminus6)timesLog(Age in quarters+1) 0038 (0033) 0040 (0031)Quarterly returns(tminus7)timesLog(Age in quarters+1) 0013 (0033) 0007 (0032)Quarterly returns(tminus8)timesLog(Age in quarters+1) 0033 (0033) 0034 (0033)Quarterly returns(tminus9)timesLog(Age in quarters+1) 0003 (0040) minus0004 (0040)Quarterly returns(tminus10)timesLog(Age in quarters+1) 0132 (0038) 0133 (0035)Quarterly returns(tminus11)timesLog(Age in quarters+1) 0018 (0040) 0007 (0041)Quarterly returns(tminus1) 1325 (0117) 1233 (0113)Quarterly returns(tminus2) 0773 (0095) 0602 (0087)Quarterly returns(tminus3) 0354 (0114) 0287 (0110)Quarterly returns(tminus4) 0124 (0112) 0110 (0106)Quarterly returns(tminus5) 0200 (0100) 0215 (0092)Quarterly returns(tminus6) minus0067 (0112) minus0056 (0104)Quarterly returns(tminus7) 0033 (0106) 0079 (0099)Quarterly returns(tminus8) minus0066 (0107) minus0046 (0107)Quarterly returns(tminus9) 0004 (0128) 0053 (0127)Quarterly returns(tminus10) minus0416 (0122) minus0383 (0114)Quarterly returns(tminus11) minus0038 (0126) 0037 (0130)Log(Age + 1) minus0026 (0007) minus0000 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0168 0198F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1788 (0000) 1570 (0000)All age interaction terms 1214 (0000) 1084 (0000)
Panel C Scalability interactions
Quarterly returns(tminus1)timesConstrained minus0215 (0048) minus0215 (0047)Quarterly returns(tminus2)timesConstrained minus0096 (0043) minus0078 (0039)Quarterly returns(tminus3)timesConstrained minus0102 (0038) minus0107 (0037)Quarterly returns(tminus4)timesConstrained 0013 (0034) 0013 (0036)Quarterly returns(tminus5)timesConstrained minus0009 (0032) minus0022 (0032)Quarterly returns(tminus6)timesConstrained 0026 (0034) 0013 (0034)Quarterly returns(tminus7)timesConstrained minus0044 (0036) minus0057 (0036)Quarterly returns(tminus8)timesConstrained 0032 (0037) 0033 (0036)Quarterly returns(tminus9)timesConstrained minus0085 (0030) minus0090 (0031)Quarterly returns(tminus10)timesConstrained minus0023 (0036) minus0007 (0036)Quarterly returns(tminus11)timesConstrained minus0007 (0037) minus0004 (0038)
(Continued)
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 883
Table IIEstimates of Flow-Performance Sensitivity
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) using quarterly data The dependent variable is Quarterly flow as defined byequation (2) See Table I for definitions of all independent variables The sample period is from 1995to 2010 We only employ the funds with a subscription period less than or equal to three months toavoid quarterly inflows being restricted by subscription periods Equation (1) is augmented in PanelB to include interaction terms of the log of the fundrsquos age plus one with prior performance and inPanel C to include interaction terms of Constrained with performance variables All specificationsin Panel A and B include style fixed effects quarter fixed effects and quarter-by-style fixed effectsOnly quarter fixed effects are included in Panel C due to multicollinearity Standard errors aredouble clustered by fund and year and indicate statistical significance at the 1 5and 10 level respectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel A Base-case
Quarterly returns(tminus1) 0501 (0045) 0496 (0043)Quarterly returns(tminus2) 0314 (0039) 0263 (0036)Quarterly returns(tminus3) 0209 (0034) 0190 (0028)Quarterly returns(tminus4) 0117 (0035) 0111 (0030)Quarterly returns(tminus5) 0089 (0030) 0098 (0027)Quarterly returns(tminus6) 0054 (0033) 0066 (0030)Quarterly returns(tminus7) 0083 (0027) 0103 (0025)Quarterly returns(tminus8) 0044 (0022) 0060 (0022)Quarterly returns(tminus9) 0012 (0029) 0034 (0029)Quarterly returns(tminus10) minus0010 (0030) 0019 (0029)Quarterly returns(tminus11) 0025 (0027) 0056 (0026)Quarterly flow(tminus1) 0158 (0008)Log(Age + 1) minus0004 (0005)Log(AUM(tminus1)+1) minus0024 (0001)Volatility minus0141 (0037)Management fees 0211 (0386)Incentive fees 0005 (0053)HWM indicator 0020 (0004)Leveraged indicator 0000 (0004)Open-to-public indicator minus0002 (0005)On-shore indicator minus0032 (0005)Log(Total redemption period +1) 0034 (0007)Log(Lock-up period +1) minus0005 (0003)Log(Subscription period +1) minus0023 (0015)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0162 0194F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1940 (0000) 2077 (0000)
(Continued)
884 The Journal of Finance Rcopy
Table IImdashContinued
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel B Age interactions
Quarterly returns(tminus1)timesLog(Age in quarters+1) minus0319 (0035) minus0285 (0034)Quarterly returns(tminus2)timesLog(Age in quarters+1) minus0171 (0027) minus0125 (0025)Quarterly returns(tminus3)timesLog(Age in quarters+1) minus0051 (0036) minus0033 (0036)Quarterly returns(tminus4)timesLog(Age in quarters+1) minus0003 (0032) 0001 (0032)Quarterly returns(tminus5)timesLog(Age in quarters+1) minus0041 (0030) minus0042 (0029)Quarterly returns(tminus6)timesLog(Age in quarters+1) 0038 (0033) 0040 (0031)Quarterly returns(tminus7)timesLog(Age in quarters+1) 0013 (0033) 0007 (0032)Quarterly returns(tminus8)timesLog(Age in quarters+1) 0033 (0033) 0034 (0033)Quarterly returns(tminus9)timesLog(Age in quarters+1) 0003 (0040) minus0004 (0040)Quarterly returns(tminus10)timesLog(Age in quarters+1) 0132 (0038) 0133 (0035)Quarterly returns(tminus11)timesLog(Age in quarters+1) 0018 (0040) 0007 (0041)Quarterly returns(tminus1) 1325 (0117) 1233 (0113)Quarterly returns(tminus2) 0773 (0095) 0602 (0087)Quarterly returns(tminus3) 0354 (0114) 0287 (0110)Quarterly returns(tminus4) 0124 (0112) 0110 (0106)Quarterly returns(tminus5) 0200 (0100) 0215 (0092)Quarterly returns(tminus6) minus0067 (0112) minus0056 (0104)Quarterly returns(tminus7) 0033 (0106) 0079 (0099)Quarterly returns(tminus8) minus0066 (0107) minus0046 (0107)Quarterly returns(tminus9) 0004 (0128) 0053 (0127)Quarterly returns(tminus10) minus0416 (0122) minus0383 (0114)Quarterly returns(tminus11) minus0038 (0126) 0037 (0130)Log(Age + 1) minus0026 (0007) minus0000 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0168 0198F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1788 (0000) 1570 (0000)All age interaction terms 1214 (0000) 1084 (0000)
Panel C Scalability interactions
Quarterly returns(tminus1)timesConstrained minus0215 (0048) minus0215 (0047)Quarterly returns(tminus2)timesConstrained minus0096 (0043) minus0078 (0039)Quarterly returns(tminus3)timesConstrained minus0102 (0038) minus0107 (0037)Quarterly returns(tminus4)timesConstrained 0013 (0034) 0013 (0036)Quarterly returns(tminus5)timesConstrained minus0009 (0032) minus0022 (0032)Quarterly returns(tminus6)timesConstrained 0026 (0034) 0013 (0034)Quarterly returns(tminus7)timesConstrained minus0044 (0036) minus0057 (0036)Quarterly returns(tminus8)timesConstrained 0032 (0037) 0033 (0036)Quarterly returns(tminus9)timesConstrained minus0085 (0030) minus0090 (0031)Quarterly returns(tminus10)timesConstrained minus0023 (0036) minus0007 (0036)Quarterly returns(tminus11)timesConstrained minus0007 (0037) minus0004 (0038)
(Continued)
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
884 The Journal of Finance Rcopy
Table IImdashContinued
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Panel B Age interactions
Quarterly returns(tminus1)timesLog(Age in quarters+1) minus0319 (0035) minus0285 (0034)Quarterly returns(tminus2)timesLog(Age in quarters+1) minus0171 (0027) minus0125 (0025)Quarterly returns(tminus3)timesLog(Age in quarters+1) minus0051 (0036) minus0033 (0036)Quarterly returns(tminus4)timesLog(Age in quarters+1) minus0003 (0032) 0001 (0032)Quarterly returns(tminus5)timesLog(Age in quarters+1) minus0041 (0030) minus0042 (0029)Quarterly returns(tminus6)timesLog(Age in quarters+1) 0038 (0033) 0040 (0031)Quarterly returns(tminus7)timesLog(Age in quarters+1) 0013 (0033) 0007 (0032)Quarterly returns(tminus8)timesLog(Age in quarters+1) 0033 (0033) 0034 (0033)Quarterly returns(tminus9)timesLog(Age in quarters+1) 0003 (0040) minus0004 (0040)Quarterly returns(tminus10)timesLog(Age in quarters+1) 0132 (0038) 0133 (0035)Quarterly returns(tminus11)timesLog(Age in quarters+1) 0018 (0040) 0007 (0041)Quarterly returns(tminus1) 1325 (0117) 1233 (0113)Quarterly returns(tminus2) 0773 (0095) 0602 (0087)Quarterly returns(tminus3) 0354 (0114) 0287 (0110)Quarterly returns(tminus4) 0124 (0112) 0110 (0106)Quarterly returns(tminus5) 0200 (0100) 0215 (0092)Quarterly returns(tminus6) minus0067 (0112) minus0056 (0104)Quarterly returns(tminus7) 0033 (0106) 0079 (0099)Quarterly returns(tminus8) minus0066 (0107) minus0046 (0107)Quarterly returns(tminus9) 0004 (0128) 0053 (0127)Quarterly returns(tminus10) minus0416 (0122) minus0383 (0114)Quarterly returns(tminus11) minus0038 (0126) 0037 (0130)Log(Age + 1) minus0026 (0007) minus0000 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 50333 50333Adjusted R2 0168 0198F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1788 (0000) 1570 (0000)All age interaction terms 1214 (0000) 1084 (0000)
Panel C Scalability interactions
Quarterly returns(tminus1)timesConstrained minus0215 (0048) minus0215 (0047)Quarterly returns(tminus2)timesConstrained minus0096 (0043) minus0078 (0039)Quarterly returns(tminus3)timesConstrained minus0102 (0038) minus0107 (0037)Quarterly returns(tminus4)timesConstrained 0013 (0034) 0013 (0036)Quarterly returns(tminus5)timesConstrained minus0009 (0032) minus0022 (0032)Quarterly returns(tminus6)timesConstrained 0026 (0034) 0013 (0034)Quarterly returns(tminus7)timesConstrained minus0044 (0036) minus0057 (0036)Quarterly returns(tminus8)timesConstrained 0032 (0037) 0033 (0036)Quarterly returns(tminus9)timesConstrained minus0085 (0030) minus0090 (0031)Quarterly returns(tminus10)timesConstrained minus0023 (0036) minus0007 (0036)Quarterly returns(tminus11)timesConstrained minus0007 (0037) minus0004 (0038)
(Continued)
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 885
Table IImdashContinued
Panel C Scalability interactions
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0559 (0049) 0554 (0046)Quarterly returns(tminus2) 0346 (0045) 0291 (0040)Quarterly returns(tminus3) 0241 (0039) 0224 (0032)Quarterly returns(tminus4) 0099 (0037) 0095 (0034)Quarterly returns(tminus5) 0077 (0034) 0094 (0030)Quarterly returns(tminus6) 0042 (0040) 0062 (0036)Quarterly returns(tminus7) 0101 (0032) 0126 (0030)Quarterly returns(tminus8) 0034 (0026) 0049 (0026)Quarterly returns(tminus9) 0034 (0033) 0060 (0033)Quarterly returns(tminus10) minus0002 (0035) 0028 (0033)Quarterly returns(tminus11) 0015 (0032) 0050 (0031)Constrained indicator 0012 (0007) 0010 (0006)Indicators for missing lagged returns Yes YesOther controls No YesFixed effects
Quarter Yes YesNumber of observations 50333 50333Adjusted R2 0156 0189F-test for joint significance of F-stat (p-value) F-stat (p-value)
All lagged returns 1980 (0000) 2065 (0000)All scalability interaction terms 476 (0000) 424 (0000)
Theoretically a learning framework such as Berk and Green (2004) suggeststhat the sensitivity of fund flows to performance should depend on the precisionof the prior distribution of ability The precision of the prior distribution islikely to be related to the experience of the fund managers Intuitively a moreexperienced manager is more of a ldquoknown quantityrdquo so given an observationof performance an observer will update her assessment of his ability less thanif the same performance were observed from a new manager In addition thesensitivity of inflows to performance should depend on the extent to which it ispossible to replicate the current distribution of returns if the fund increases insize in other words the fund strategyrsquos ldquoscalabilityrdquo
To evaluate these implications we estimate the extent to which the sensitiv-ity of inflows to performance depends on the fundrsquos age To do so we estimateequation (1) including interaction terms of prior performance plus the log ofone plus the fundrsquos age and present these estimates in Panel B of Table IIIn each estimated equation the sum of the interaction coefficients is negativeand the coefficients are jointly statistically significant with a majority of theeffect arising from the two quarters immediately preceding the focal quarterThe negative coefficients on the interaction terms mean that as hedge fundsget older the effect of performance on inflows declines
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
886 The Journal of Finance Rcopy
A fundrsquos strategy likely affects the sensitivity of inflows to performance be-cause some strategies can be replicated with more capital while others willface diminishing returns For example arbitrage strategies (eg Convert-ible Arbitrage) in which opportunities disappear as they are exploited areunlikely to be infinitely scalable by nature Strategies that invest in illiq-uid assets and have high market impact costs (eg Event Driven) are alsomore likely to face capacity constraints (Aragon (2007) Teo (2009) Getmansky(2012)) However strategies that involve liquid instruments (eg LongShortEquity Equity Market Neutral) are less prone to capacity constraints Ra-madorai (2013) finds a negative effect of capacity constraints on hedge fundreturns
To evaluate whether scalability affects the flow-performance relation werely on the classification of Ding et al (2009) who consider Convertible Arbi-trage Emerging Market Event Driven and Fixed Income Arbitrage strategiesto be ldquocapacity constrainedrdquo The other strategies (Equity Market NeutralGlobal Macro LongShort Equity Multi-Strategy and Others) are classified asldquounconstrainedrdquo We create an indicator variable equal to one if the fund iscapacity constrained and zero if it is unconstrained We interact this variablewith the past performance variables and present the results in Panel C ofTable II
As with the previous estimates the coefficients on lagged performance arepositive and statistically significantly different from zero However the coeffi-cients on these variables interacted with the ldquoConstrainedrdquo indicator variableare negative and statistically significant for the three most recent quartersimplying that the strategies we consider to be constrained are less respon-sive in size to a performance shock Even though a shock to performance forldquoconstrainedrdquo funds would cause the market to update its assessment of fundmanagersrsquo abilities the fact that they are less scalable limits the extent towhich investors are willing to change their investments in these funds as aresult
A caveat to these results is that if hedge funds misreport their returnsestimates of the flow-performance relation may be biased In particular Bollenand Pool (2009) show that the potential for misreporting is strongest whenreturns are slightly negative In this case the relation between flows and trueperformance would be weaker than the relation between flows and performancereported to the database leading our estimates of indirect incentives (withrespect to true performance) to be overstated To gauge the magnitude of thiseffect we reestimate the flow-performance equations under the assumptionthat all reported positive returns were accurate but all negative returns werein fact upward-biased by 10 The resulting coefficients are about 26 lowerthan those reported above which suggests that misreporting may lead to abouta 26 overstatement of the part of indirect incentives (reported below) thatcomes from new inflows
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 887
IV Calculating Indirect and Direct Pay for Performance
In this section we use the models discussed in Section I and the Appendicestogether with the estimates presented in Section III to quantify the magni-tude of direct and indirect pay for performance sensitivities facing hedge fundmanagers
To calculate direct pay for performance we follow Agarwal Daniel andNaikrsquos (2009) total delta approach using the parameters discussed in AppendixA This approach takes the perspective of a manager at the beginning of the pe-riod calculating the sensitivity of his claim to the fundrsquos assets to performancerealized over the upcoming period Based on a set of assumptions discussedin Appendix A we calculate direct incentives arising from each individual in-vestorrsquos assets as well as the managerrsquos personal stake and then sum themup to obtain the total direct pay for performance for each fund-quarter in oursample We take the average of these fund-quarter estimates to be our estimateof typical direct pay for performance sensitivities
For indirect pay for performance the coefficients on returns in Table II carrythe interpretation of incremental inflows as a percentage of beginning-of-periodAUM for an incremental percentage point of returns As previously discussedwe use four models to estimate the fraction of an incremental dollar investedin the fund that accrues to the manager in expected future management andincentive fees the base GIR model the GIR model augmented to account for fu-ture performance-based flows the LWY model (which incorporates endogenousleverage) and the LWY model extended to account for future performance-based flows For every fund-quarter we use each of the four models to calculatethe fraction of each incremental dollar that in expectation will accrue to man-agers Then as with direct pay for performance we take the average of thefund-quarter estimates to be our estimate of typical indirect pay for perfor-mance sensitivities We repeat this calculation for a number of alternativeparameter choices Details of the models parameter choices and calculationsare provided in Appendix B
We use common parameters for each model to ensure an apples-to-applescomparison across models For each of the four models we calculate indirectincentives using eight sets of parameters obtained from two choices of each ofthree parameters The three parameters we vary are those that because of theireconomic interpretation are likely to have a quantitatively important impacton our estimates of indirect incentives A particularly important parameter isb which represents the minimum asset value relative to the HWM that theinvestor will tolerate before withdrawing all of his or her money from the fundIf b = 0 the fund is never liquidated for poor performance Positive values of bimply a positive probability of performance-related liquidation each period andtherefore a finite expected fund life We consider b = 08 as recommended byGIR which means that a 20 loss results in liquidation of an investorrsquos stakeWe also present estimates using b = 0685 which LWY use in their analysis
Another important parameter is δ + λ which is the fraction of an investorrsquoscapital that he or she withdraws each period for exogenous reasons GIR set
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
888 The Journal of Finance Rcopy
this parameter equal to 5 per year which corresponds to the typical spendingrules of institutional investors such as endowments or pension plans LWYuse 10 which although too high for such institutions may be appropriatefor other investor types The higher this parameter the lower the indirectincentives because higher withdrawal rates mean that new money will stayin the fund for a shorter period of time We present estimates using quarterlyequivalents of 5 and 10
The final parameter that we vary is the managerrsquos future expected gross-of-fee risk-adjusted performance α Following GIR we present estimates forquarterly equivalents of α = 0 and 3 which correspond to levered valuesin the LWY framework LWY calibrate their model to an unlevered α = 122which is close to a levered value of 3 given a typical hedge fund leverage
In both GIR and LWY α is exogenous and time-invariant In particular itis not related to current or future inflows of new investment If inflows lead tolower future performance our estimates of indirect pay for performance wouldbe overstated Although Naik Ramadorai and Stromqvist (2007) AgrawalDaniel and Naik (2009) and Fung et al (2008) do find that inflows result inlower future performance this relation is not apparent in our more recent dataMoreover given the magnitude of such relations identified by prior work anyoverstatement of indirect pay for performance is likely to be small For instanceour estimates of the flow-past performance sensitivity presented in Panel Aof Table II imply that a one-percentage-point incremental return in a givenquarter leads to a 15 increase in AUM over the next 11 quarters Accountingfor the magnitude of the deleterious effects of flows on future performanceestimated by Agrawal Daniel and Naik (2009) would reduce this estimate byonly 001 of AUM
A Estimates of Direct and Indirect Incentives
The estimates of direct incentives are summarized in Panel A of Table IIIThese calculations indicate that the expected dollar increase in incentive feesfor an incremental one-percentage-point increase in quarterly net return equals$142000 (Row (1)) or roughly 68 cents for every dollar returned to investors(Row (4)) This figure is lower than the typical 20 incentive fee rate becausethe option is always out of the money (when the option is in the money theincentive fee is immediately paid and the strike is reset) The change in man-agersrsquo personal stakes averages $190000 for an incremental 1 return (Row(2)) and 95 cents for every dollar returned to investors (Row (5)) Total directincentives average $331000 for a one-percentage-point increase in fund value(Row (3)) or 163 cents for each additional dollar created (Row (6))
Panels B through E of Table III perform comparable calculations for indirectincentives using each of the four models to value the fundrsquos future fee incomeFor example the estimates in Panel B are from the base-case GIR model thatdoes not allow for endogenous leverage or future performance-based fund flowsConsider first the estimates reported in Column (1) (with b = 0685 α = 0 andδ + λ = 5) These estimates indicate that for a one-percentage-point increase
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 889
Tab
leII
ID
irec
tan
dIn
dir
ect
Pay
for
Per
form
ance
Th
ista
ble
pres
ents
esti
mat
esof
dire
ctan
din
dire
ctpa
yfo
rpe
rfor
man
ce(i
nce
nti
ves)
ata
quar
terl
yfr
equ
ency
A
llre
port
edst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
inth
eda
taP
anel
Apr
esen
tses
tim
ates
ofdi
rect
ince
nti
ves
calc
ula
ted
asde
scri
bed
inA
ppen
dix
AD
irec
tin
cen
tive
sar
ede
fin
edas
the
expe
cted
pres
ent
valu
edo
llar
chan
gein
man
ager
wea
lth
from
dire
ctin
cen
tive
fees
plu
sth
em
anag
errsquos
own
ersh
ipst
ake
resu
ltin
gfr
oman
incr
emen
tal
one-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
fun
dre
turn
sP
anel
sB
toE
pres
ent
esti
mat
esof
indi
rect
ince
nti
ves
calc
ula
ted
usi
ng
fou
rdi
ffer
ent
mod
els
asde
scri
bed
inA
ppen
dix
BI
ndi
rect
ince
nti
ves
are
defi
ned
asth
eex
pect
edpr
esen
tva
lue
doll
arch
ange
inm
anag
erw
ealt
hfr
omfu
ture
fees
earn
edfr
omin
flow
sof
new
inve
stm
ent
plu
sth
ein
crea
sein
the
valu
eof
exis
tin
gas
sets
resu
ltin
gfr
oman
incr
emen
talo
ne-
perc
enta
ge-p
oin
tor
one-
doll
arin
crea
sein
retu
rns
toin
vest
ors
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
A
llpa
ram
eter
choi
ces
are
prov
ided
inTa
ble
CI
inA
ppen
dix
C
Th
en
um
ber
offu
nd-
quar
ters
use
din
all
colu
mn
sof
Pan
els
Ato
Cis
497
76(5
57fu
nd-
quar
ter
obse
rvat
ion
sar
edr
oppe
dfr
omth
efu
llre
gres
sion
sam
ple
beca
use
the
sum
ofth
eva
lues
ofal
lin
vest
orsrsquo
stak
eis
zero
for
thes
eca
ses
acco
rdin
gto
the
com
puta
tion
sde
scri
bed
inA
ppen
dix
A)
Th
en
um
ber
ofob
serv
atio
ns
use
dto
esti
mat
eth
eLW
Ym
odel
(Pan
elD
)an
dth
eex
ten
ded
LWY
mod
el(P
anel
E)
are
som
ewh
atsm
alle
rbe
cau
seth
eO
DE
ineq
uat
ion
(B3)
fail
sto
hav
ea
nu
mer
ical
solu
tion
for
cert
ain
com
bin
atio
ns
ofpa
ram
eter
sT
he
nu
mbe
rof
fun
d-qu
arte
rsu
sed
ines
tim
atio
nav
erag
es49
731
inP
anel
Dan
d48
033
inP
anel
E
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Dir
ect
effe
ctfr
omin
cen
tive
fees
014
20
142
014
20
142
014
20
142
014
20
142
(2)
Dir
ect
effe
ctfr
omm
anag
ersrsquo
pers
onal
stak
e0
190
019
00
190
019
00
190
019
00
190
019
0(3
)To
tald
irec
tef
fect
033
10
331
033
10
331
033
10
331
033
10
331
Per
1$ch
ange
infu
nd
valu
e($
)(4
)D
irec
tef
fect
from
ince
nti
vefe
es0
068
006
80
068
006
80
068
006
80
068
006
8(5
)D
irec
tef
fect
from
man
ager
srsquope
rson
alst
ake
009
50
095
009
50
095
009
50
095
009
50
095
(6)
Tota
ldir
ect
effe
ct0
163
016
30
163
016
30
163
016
30
163
016
3
(Con
tin
ued
)
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
890 The Journal of Finance Rcopy
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Goe
tzm
ann
In
gers
oll
and
Ros
s(G
IR2
003)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
048
20
355
109
40
589
030
80
248
068
90
436
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
449
033
90
903
051
70
340
028
20
671
044
9(3
)To
tali
ndi
rect
effe
ct0
932
069
51
998
110
60
648
053
11
360
088
4(4
)In
dire
ctd
irec
t2
812
106
033
341
961
604
112
67P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
022
80
169
048
80
272
014
50
116
030
50
195
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
193
014
60
383
022
20
138
011
40
269
018
0(7
)To
tali
ndi
rect
effe
ct0
421
031
50
871
049
40
283
023
00
574
037
5(8
)In
dire
ctd
irec
t2
581
935
353
031
741
413
522
30
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfl
ows
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
054
50
396
113
50
622
034
60
279
070
10
455
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
474
035
50
892
051
80
354
029
40
642
044
0(3
)To
tali
ndi
rect
effe
ct1
018
075
12
028
114
00
700
057
31
343
089
5(4
)In
dire
ctd
irec
t3
072
276
123
442
111
734
052
70P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
189
050
90
289
016
20
130
031
20
205
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
208
015
60
386
022
80
147
012
20
263
018
0(7
)To
tali
ndi
rect
effe
ct0
465
034
50
895
051
70
309
025
20
575
038
6(8
)In
dire
ctd
irec
t2
852
125
503
171
901
553
532
37
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 891
Tab
leII
ImdashC
onti
nu
ed
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
Lan
Wan
gan
dYa
ng
(LW
Y20
13)
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
133
70
691
042
50
315
072
50
454
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
11
245
067
30
551
038
70
821
052
2(3
)To
tali
ndi
rect
effe
ct1
137
072
72
582
136
40
976
070
21
546
097
6(4
)In
dire
ctd
irec
t3
432
197
804
122
952
124
672
95P
er1$
chan
gein
fun
dva
lue
($)
(5)
Indi
rect
effe
ctfr
omn
ewm
oney
025
70
169
057
50
305
019
80
146
032
10
204
(6)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
241
015
10
498
027
20
217
015
30
321
020
5(7
)To
tali
ndi
rect
effe
ct0
498
032
01
073
057
70
415
030
00
642
040
8(8
)In
dire
ctd
irec
t3
051
966
583
542
551
843
942
51
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Indi
rect
effe
ctfr
omn
ewm
oney
055
40
366
214
41
148
042
50
315
092
50
545
(2)
Indi
rect
effe
ctfr
omch
ange
inva
lue
ofex
isti
ng
asse
ts0
583
036
12
126
115
20
551
038
71
053
063
7(3
)To
tali
ndi
rect
effe
ct1
137
072
74
270
230
00
976
070
21
978
118
2(4
)In
dire
ctd
irec
t3
432
1912
89
694
295
212
597
357
Per
1$ch
ange
infu
nd
valu
e($
)(5
)In
dire
ctef
fect
from
new
mon
ey0
257
016
90
938
049
50
198
014
60
402
024
0(6
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
024
10
151
085
30
453
021
70
153
040
40
246
(7)
Tota
lin
dire
ctef
fect
049
80
320
179
10
948
041
50
300
080
50
487
(8)
Indi
rect
dir
ect
305
196
110
05
822
551
844
942
99
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
892 The Journal of Finance Rcopy
in returns the incremental future fees from inflows of new investment average$482000 and the incremental future fees from the increase in value of existingassets average $449000 The total is $932000 which is 281 times the directchange in compensation through incentive fees and the managersrsquo personalstakes Expressed as a fraction of an incremental dollar in the fund managersreceive 228 cents from new money and 193 cents from the change in the valueof existing assets Together these effects are 258 times the direct incentivesthat managers receive for the same change14
The remaining columns of Panel B present analogous calculations for al-ternative parameter choices In general higher choices of b and δ + λ reducethe size of indirect incentives while a higher α raises them Intuitively ahigher b implies a lower tolerance for negative return shocks by investors soin expectation future fees and hence indirect incentives will be lower whenb is higher Similarly a higher value of δ + λ means that assets exit the fundmore quickly for nonperformance-related reasons which effectively shortensa fundrsquos expected life so future fees will be lower In contrast a higher αmeans that future returns are expected to be higher so the likelihood of hittingthe liquidation boundary defined by b declines and fees will be a percentageof higher asset values so their expectation increases with α Nonetheless re-gardless of the choice of parameters indirect incentives are substantially largerthan direct incentives with indirect-to-direct incentive ratios varying from 14to 60
Considering the other models presented in Panels C D and E it is evi-dent that for any set of parameters the indirect incentives coming from thealternative models are higher than those from the base-case GIR model Theaugmented GIR model and the two LWY models differ from the base-case GIRmodel by accounting for future performance-related fund flows andor portfolioallocations that are chosen endogenously to incorporate managersrsquo incentives
In the case of the augmented GIR model future performance-based flows areequivalent to an increase in fund volatility Volatility increases the value of themanagerrsquos claim to each incremental dollar in the fund because the managerrsquosincentive fees are options on the fund15 Indirect incentives calculated usingthe augmented GIR model and presented in Panel C are about 9 larger thanfor the base GIR model when α = 0 and are about 2 larger when α = 3
The LWY model allows the manager to choose the leverage of the fund inresponse to incentives When managers can endogenously adjust their futureportfolios they will do so only when it benefits them This effect increases thevalue to the manager of an incremental dollar in the fund compared to when
14 The indirect-to-direct incentives ratios are calculated by dividing average indirect incentivesby average direct incentives for both ldquoper incremental percentage point of returnsrdquo and ldquoper incre-mental dollarrdquo calculations The two ratios are not the same because for each fund-quarter directand indirect incentives per 1 increase in returns are scaled by (1 of) AUM which varies overtime to reach direct and indirect incentives per $1 change to fund value
15 In the GIR framework the standard intuition that volatility increases option value is partiallybut not entirely offset by the fact that greater volatility increases the probability of liquidationgiven a fixed b
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 893
they do not have this option Indirect incentives calculated using the LWYmodel and presented in Panel D range from 2 to 51 larger than in the baseGIR model depending on the model parameters
The LWY model augmented for performance-based fund flows accounts notonly for endogenous allocation but also inflows of new investment in the futureif future performance is good Both channels increase the value to the managerof an incremental dollar in the fund They also interact in that with the possi-bility of future inflows the manager endogenously takes more risk This effectis especially strong when large α is combined with low b as the downside fromrisk-taking (hitting the liquidation boundary represented by b) is less likely tooccur in this case Thus indirect incentives from the augmented LWY modelpresented in Panel E are about 39 higher than in the base-case GIR modelwhen the liquidation threshold is tight (b = 08) When the liquidation thresh-old is looser (b = 0685) indirect incentives are about 12 higher comparedto the base GIR model for unskilled managers (α = 0) and 105 higher forskilled ones (α = 3)
Overall the ratio of indirect to direct incentives varies from 14 to 129 acrossthe 32 model and parameter combinations examined here with an average of35 The three alternative models typically deliver estimates of indirect in-centives that are roughly 25 higher than the base-case GIR model becauseperformance-based inflows and endogenous portfolios both increase future pay-ments to fund managers Regardless of the specific calculations used it is ev-ident that indirect incentives for most hedge fund managers are considerablylarger than their direct incentives16
B Direct and Indirect Incentives by Fund Age
Panel B of Table II documents that the magnitude of the flow-performancerelation declines as a fund ages consistent with the logic of Berk and Green(2004) Because the indirect incentives that managers face are due in largepart to the influence of returns on flows it is likely that managersrsquo indirectincentives also decline with a fundrsquos age
Table IV examines this hypothesis by calculating indirect incentives for fundsof different ages with parameter choices given in Column (4) of Table III (ieb = 0685 α = 3 and δ + λ = 10 these are the parameter choices used byLWY) Here we focus our discussion on changes in manager wealth per in-cremental dollar returned to investors because older funds are systematicallylarger complicating comparisons across age groups of changes in wealth perincremental percentage point of returns However for completeness we reportthe latter incentive measure as well in Table IV
Panel A presents the average direct pay for performance for funds of differentages Direct pay for performance appears to increase with fund age For a new
16 Indirect incentives are also large in the case in which the manager has the option to resetthe HWM by strategically abandoning the fund and starting a new one This case can also beaccommodated in the LWY framework The estimates of indirect incentives would fall in betweenthose for the base LWY model and the LWY model augmented for performance-based flows
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
894 The Journal of Finance Rcopy
Tab
leIV
Dir
ect
and
Ind
irec
tP
ayfo
rP
erfo
rman
ceb
yA
geG
rou
pT
his
tabl
epr
esen
tses
tim
ates
ofdi
rect
and
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)at
aqu
arte
rly
freq
uen
cys
imil
arto
Tabl
eII
Ibu
tbr
oken
dow
nby
fun
dag
eR
epor
ted
stat
isti
csar
eav
erag
esta
ken
over
all
fun
dsof
agi
ven
age
All
esti
mat
esin
this
tabl
eu
seth
equ
arte
rly
equ
ival
ent
ofth
epa
ram
eter
sb
=0
685α
=3
an
dδ+λ
=10
t
hes
ear
eth
epa
ram
eter
sch
osen
byLW
Y
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elA
Dir
ect
ince
nti
ves
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
ldir
ect
effe
ct0
100
150
180
230
290
380
460
510
580
720
851
031
321
281
221
04P
er1$
chan
gein
fun
dva
lue
($)
(2)
Tota
ldir
ect
effe
ct0
100
110
120
140
160
180
210
230
240
260
280
290
310
320
360
35
Pan
elB
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
el
Per
1C
han
gein
fun
dva
lue
($M
)(1
)To
tali
ndi
rect
effe
ct0
590
820
940
981
081
121
221
311
401
631
812
042
272
261
961
50(2
)In
dire
ctd
irec
t6
055
435
134
273
752
982
632
572
402
272
131
981
721
771
611
44P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
057
053
050
048
047
045
044
043
042
042
042
042
042
042
042
039
(2)
Indi
rect
dir
ect
551
478
407
333
290
242
209
187
172
159
150
146
134
130
117
110
Pan
elC
In
dire
ctin
cen
tive
ses
tim
ated
from
the
GIR
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
062
084
096
100
110
117
127
134
143
171
186
207
233
234
207
153
(2)
Indi
rect
dir
ect
631
560
524
435
385
311
273
261
245
239
219
201
177
183
170
147
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
590
550
530
500
490
470
460
450
440
440
440
440
430
440
440
40(2
)In
dire
ctd
irec
t5
765
004
253
483
032
542
191
961
801
681
561
511
391
361
231
13
(Con
tin
ued
)
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 895
Tab
leIV
mdashC
onti
nu
ed
Fu
nd
age
(yea
rs)
01
23
45
67
89
1011
1213
14
15
Pan
elD
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
el
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
071
100
116
119
133
140
154
172
186
203
225
253
288
262
228
172
(2)
Indi
rect
dir
ect
728
666
633
515
466
371
332
336
319
283
264
246
218
204
187
166
Per
1$ch
ange
infu
nd
valu
e($
)(1
)To
tali
ndi
rect
effe
ct0
650
610
590
560
550
530
520
520
510
500
500
510
500
500
500
45(2
)In
dire
ctd
irec
t6
305
504
743
883
432
882
472
262
071
891
771
741
601
541
411
27
Pan
elE
In
dire
ctin
cen
tive
ses
tim
ated
from
the
LWY
mod
elau
gmen
ted
for
perf
orm
ance
-bas
edfu
nd
flow
s
Per
1ch
ange
infu
nd
valu
e($
M)
(1)
Tota
lin
dire
ctef
fect
118
167
191
194
224
235
263
302
327
347
386
434
500
442
382
285
(2)
Indi
rect
dir
ect
120
811
16
104
98
447
816
265
685
915
614
854
544
213
793
453
132
74P
er1$
chan
gein
fun
dva
lue
($)
(1)
Tota
lin
dire
ctef
fect
106
098
095
091
091
088
086
087
086
084
084
087
085
085
087
075
(2)
Indi
rect
dir
ect
102
68
907
706
305
674
794
113
813
503
182
982
992
742
642
452
14
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
896 The Journal of Finance Rcopy
fund an incremental dollar returned to fund investors results in 10 cents inincremental incentive fees and returns on the managerrsquos personal stake Fora 10-year-old fund this quantity increases to 28 cents Much of this increasehowever occurs because of our assumption that managers reinvest their feesin which case their ownership increases over time To the extent that managersstart their funds with a positive stake and do not reinvest all of their fees directpay for performance will not increase as fast as suggested by this table
Panels B to E present estimates of indirect incentives using each of the fourmodels to evaluate the fraction of fund value going to managers in the form offuture fees The estimates for each model imply that indirect incentives declinesubstantially as a fund ages For new funds an incremental dollar returned toinvestors today results in at least $057 in expected future fees This indirecteffect is about six times as important as the direct effect of performance oncurrent income from incentive fees and gains on the managerrsquos own stake Theindirect effect declines sharply with the fundrsquos age but remains larger thanthe direct effect even for 15-year-old funds
C Direct and Indirect Incentives and Fund Scalability
Indirect incentives result from managers being able to increase their fundsrsquosize when performance has been good Their ability to do so depends on theability of managers to scale their investments Some funds are relatively un-constrained in that they adopt strategies that are scalable implying that thefund can likely invest new capital with the same ex ante return distributionas existing capital In contrast other more constrained funds typically cannotaccept more capital without significantly reducing expected returns
Table V provides estimates of indirect and direct incentives for funds clas-sified as ldquoNot Capacity Constrainedrdquo and ldquoCapacity Constrainedrdquo accordingto the classifications discussed in Section IIIB The direct incentives of eachtype are very similar 16 cents for each incremental dollar returned to in-vestors for unconstrained funds compared to 17 cents for constrained fundsThe differences in indirect incentives however are substantial 52 cents foreach dollar returned to investors for unconstrained funds compared to 43 centsfor constrained funds using the GIR model This difference implies an aver-age indirect-to-direct incentives ratio of about 32 for the unconstrained fundscompared to 25 for the constrained funds Using other models the indirect-to-direct incentives ratio ranges between 34 and 73 for the unconstrainedfunds and between 26 and 60 for the constrained funds Although our proxyfor fund scalability is imperfect these differences suggest that indirect incen-tives are relatively more important in funds adopting more scalable investmentstrategies
The cross-sectional differences in indirect incentives by fund scalability pointto broader issues that are likely to affect the scalability of the hedge fundindustry and hence indirect incentives as a whole In particular current trendsin industry growth and regulation suggest that it is possible that managers willbe less able to effectively deploy additional capital in the future than has been
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 897
Table VDirect and Indirect Pay for Performance by Scalability
This table presents estimates of direct and indirect pay for performance (incentives) at a quarterlyfrequency similar to Table III but broken down by the scalability of the fundrsquos strategy ldquoCapacityconstrainedrdquo strategies are Emerging Market Fixed Income Arbitrage Event Driven and Con-vertible Arbitrage All other strategies LongShort Equity Equity Market Neutral Global MacroMulti-Strategy and Others are not considered capacity constrained Reported statistics are aver-ages taken over all fund-quarters within a grouping All estimates in this table use the quarterlyequivalent of the parameters b = 0685 α = 3 and δ+λ = 10 these are the parameters chosenby LWY
Not Capacity Constrained Capacity Constrained
Panel A Direct incentives
Per 1 change in fund value ($M)(1) Total direct effect 0340 0314Per 1$ change in fund value ($)(2) Total direct effect 0160 0169
Panel B Indirect incentives estimated from the GIR model
Per 1 change in fund value ($M)(1) Total indirect effect 1213 0862(2) Indirectdirect 357 275Per 1$ change in fund value ($)(1) Total indirect effect 0516 0427(2) Indirectdirect 323 253
Panel C Indirect incentives estimated from the GIR model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 1251 0888(2) Indirectdirect 368 283Per 1$ change in fund value ($)(1) Total indirect effect 0541 0446(2) Indirectdirect 338 264
Panel D Indirect incentives estimated from the LWY model
Per 1 change in fund value ($M)(1) Total indirect effect 1475 1100(2) Indirectdirect 434 351Per 1$ change in fund value ($)(1) Total indirect effect 0597 0509(2) Indirectdirect 374 301
Panel E Indirect incentives estimated from the LWY model augmented forperformance-based fund flows
Per 1 change in fund value ($M)(1) Total indirect effect 2468 1887(2) Indirectdirect 726 602Per 1$ change in fund value ($)(1) Total indirect effect 0971 0854(2) Indirectdirect 608 505
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
898 The Journal of Finance Rcopy
Table VIDirect and Indirect Pay for Performance by Presence
of a High-Water MarkThis table presents estimates of direct and indirect pay for performance broken down by thepresence of an HWM provision Reported statistics are averages taken over all funds in a givencategory All estimates of indirect incentives are obtained from the base LWY model with parametervalues b = 0685 α = 3 and δ+λ = 10
(1) (2) (3)All Funds Funds with
funds with HWM no HWM
(1) Number of observations used in estimation 49776 35549 14227(2) AUM ($ million) 213 239 148Pay for performance per 1 change in fund value ($M)(3) Direct effect from incentive fees 0142 0132 0165(4) Direct effect from managersrsquo personal stake 0190 0188 0193(5) Total direct effect 0331 0320 0358(6) Indirect effect from new money 0691 0775 0483(7) Indirect effect from change in value of
existing assets0673 0813 0323
(8) Total indirect effect 1364 1587 0805(9) Indirectdirect 412 496 225Pay for performance per 1$ change in fund value ($)(10) Direct effect from incentive fees 0068 0054 0105(11) Direct effect from managersrsquo personal stake 0095 0074 0146(12) Total direct effect 0163 0128 0251(13) Indirect effect from new money 0305 0310 0291(14) Indirect effect from change in value of
existing assets0272 0303 0195
(15) Total indirect effect 0577 0613 0486(16) Indirectdirect 354 480 194
true in the past (ie in our sample period) If so the flow-performance relationwill decline and so will the magnitude of indirect incentives All of our estimatesshould be interpreted with this caveat in mind
D Direct and Indirect Incentives and High-Water Marks
The importance of indirect incentives could vary across funds depending onthe presence of an HWM provision because of both potential differences inflow-performance sensitivities and the impact of the HWM on fees
Table VI presents estimates of direct and indirect incentives for funds withand without HWM provisions We find that funds without HWM provisionshave higher direct incentives while funds with HWM provisions have higherindirect incentives Consequently funds with HWM provisions have substan-tially higher indirect-to-direct incentives ratios For an incremental dollarreturn to the fundrsquos investors the indirect-to-direct ratio is 194 for funds withno HWM provision and 480 for those with HWM and the difference betweenthe two is statistically significant
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 899
The fact that indirect incentives are substantially higher for funds withHWM complements the finding of Patton Ramadorai and Streatfield (2015)that the direction and magnitude of return misreporting is larger for funds withan HWM provision One reason for this finding could be the higher indirectincentives we document which suggest greater gains from misreporting
V Implications of These Estimates of Indirect Incentives
The existence of large indirect incentives for hedge fund managers has im-plications for our understanding of a number of issues including the structureof contracts in hedge funds the portfolio choices of investors and contractingbeyond the asset management context In this section we highlight these im-plications their empirical predictions as well as a number of questions thatemerge from our analysis that are potential topics for future research
A Implications for Hedge Fund Contracting and Managersrsquo Total Incentives
Hedge fund contracts are set up in a sophisticated manner with a fee struc-ture containing a combination of management fees and incentives that useHWMs that adjust their effective strike price Presumably these contracts aredesigned to solve a principal-agent problem of some sort Yet as emphasizedby Gibbons and Murphy (1992) what matters in a principal-agent problem isthe total pay for performance that agents receive not the distribution betweenits direct and indirect components
A1 Total Incentives over Time and the Relation between Directand Indirect Incentives
In the hedge fund industry our finding that indirect incentives decrease withfund age together with established evidence that explicit contracts tend to besticky implies that managersrsquo total incentives decline as a fund ages17 Whetherit is optimal for total incentives to decline and for contracts to be sticky givendeclining indirect incentives is an interesting question for future researchIf as seems likely attrition in the industry results in more experienced man-agers being higher quality on average than novice managers a theory ratio-nalizing these results would need to explain why the process of generatinghedge fund returns is such that better managers should receive weaker totalincentives
Although direct incentives in the hedge fund industry do not (given the stick-iness of the contracts) appear to increase to compensate for declining indirect
17 Funds do at times change their fee structure but such changes are infrequent Deuskar et al(2012) report a number of findings about the factors that lead hedge funds to change their feesNone of the findings in Deuskar et al (2012) suggest that changes in the hedge fund fee structureare driven by a desire to offset changes in indirect incentives faced by fund managers
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
900 The Journal of Finance Rcopy
Table VIIThe Relation between Direct and Indirect Incentives
This table presents the relations between various contractual terms and Constrained OLS coeffi-cient estimates and corresponding standard errors (in parentheses) are reported The dependentvariable is the incentive fee rate and HWM indicator in Columns (1) and (2) respectively Allanalyses are conducted at the fund level Standard errors are clustered by fund inception year and indicate statistical significance at the 1 5 and 10 level respectively
(1) (2)Incentive fee rate HWM indicator
Dependent var= Coef (SE) Coef (SE)
Constrained minus00020 (0002) minus00394 (0025)Log(Inception AUM+1) 00008 (0000) 00121 (0009)Leveraged indicator 00067 (0001) minus00136 (0027)Open-to-public indicator 00024 (0001) 01287 (0031)On-shore indicator 00045 (0002) minus00571 (0019)Log(Total redemption +1) minus00019 (0003) 00163 (0038)Log(Lock-up period+1) 00003 (0001) 01491 (0028)Log(Subscription period+1) 00194 (0006) minus04330 (0099)Number of observations 2998 2998Adjusted R2 0016 0072
incentives as a fund ages it is possible that contracts signed at fund inceptionare related to the strength of the indirect incentives that managers face In par-ticular because funds with less scalable investment strategies have a weakerflow-performance relation holding contractual terms fixed they have weakerindirect and total incentives than funds with more scalable strategies So wemight expect direct contractual incentives to be stronger in less scalable fundsto offset the weaker flow-performance relation Alternatively it is possible thatthe type of manager who manages a less scalable fund is more likely to be onewhose total incentives should be weaker In that case constrained funds couldhave weaker direct incentives than unconstrained funds
To evaluate the extent to which contracting adjusts in the presence of in-direct incentives we measure the relation between direct incentives and thescalability of an fundrsquos strategy Direct incentives are a function of the incentivefee rate which usually remains constant over the life of the fund In additionthe presence of an HWM provision mitigates a fundrsquos direct incentives to someextent Therefore to assess whether being constrained in its ability to acceptcapital because of a less scalable strategy affects the direct incentives that afund provides its managers we estimate equations predicting both the fundrsquosincentive fee rate and whether a fund employs an HWM In these equationswe include the indicator variable Constrained that equals one if a firm followsa less scalable strategy using the Ding et al (2009) classification of strategiesas well as other variables that are likely to be associated with a firmrsquos directincentives
We present estimates of these equations in Table VII In neither equa-tion is the estimated coefficient on Constrained statistically significant or of
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 901
meaningful magnitude Consequently these estimates suggest that funds donot adjust their direct compensation plans based on the extent to which theirmanagers receive indirect incentives
The lack of a relation between a hedge fundrsquos contractual fee structure andindirect market-based incentives reflects a larger puzzle about the structureof compensation in alternative asset classes Both hedge funds and privateequity funds generally use a combination of management fees that are pro-portional to capital under management and performance fees which are usu-ally 20 of profits regardless of other factors that ostensibly should affectthe magnitude of the profit-sharing percentage Although the performancefee differs more in the hedge fund industry than in private equity18 the factthat hedge funds do not appear to adjust their incentive fees based on indi-rect incentives that are large in magnitude and have wide variation acrossfunds and over time is difficult to reconcile with extant optimal contractingtheories
A2 Hedge Fund Managersrsquo Careers and Adverse Consequences of Incentives
The estimates provided above suggest that hedge fund managersrsquo careerpatterns should be heavily influenced by the returns earned by the fund whilethey are young and are establishing a reputation Since young fund managershave limited liability and the option of leaving the industry the estimatesdiscussed above imply that young hedge fund managers receive extremely largepayoffs for good performance but do not fully bear the downside risk from anyactions they take Therefore we expect the pay of managers who earn highreturns to increase substantially while there should be a high exit rate frommanagers who do not perform well
These incentives potentially lead to some adverse outcomes for investors Theexit option will lead managers to become prone to increase the risk of the fundeven if doing so comes at a cost to investors In addition if young fund managershave the option of taking investments that increase short-term returns at theexpense of the long term as suggested by Stein (1989) their incentives to doso are particularly large Thus we expect young managers to be more likely totake investments that have payoffs that are observable over shorter horizonsthan older managers In addition to the extent that accounting manipulationor fraud can increase short-term payoffs with any potential cost occurring inthe future we expect to observe such manipulation or fraud more frequentlyfrom younger managers The extent to which indirect incentives lead to theseactivities that adversely affect investors in practice is unknown
18 Gompers and Lerner (1999) Metrick and Yasuda (2010) and Robinson and Sensoy (2013)provide evidence on private equity fund compensation
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
902 The Journal of Finance Rcopy
B Implications for Investors
B1 Differences in Hedge Fund Managersrsquo Incentives
Institutional investors often highlight the importance of the incentives oftheir fund managers in making their portfolio decisions (see Swenson (2000))Since indirect incentives are such an important part of hedge fund managersrsquototal incentives they should be taken into account by investors investing inhedge funds and other kinds of assets Higher indirect incentives increase thepayoff to managers from increasing returns but also potentially lead them toincrease risk and manipulate returns The estimates presented above suggestthat younger funds and funds in more scalable sectors have higher indirectand total incentives Portfolio theory is not clear on the optimal composition ofa portfolio of funds whose incentives are different from one another Howeverit does seem clear that the differences in indirect incentives we document arefactors that a portfolio manager would like to be aware of when making portfolioallocations to hedge funds
B2 Differences in Incentives across Asset Classes
Indirect incentives are present not just for managers of hedge funds but alsofor managers of all asset classes Consequently total incentivesmdashincludingboth direct and indirect componentsmdashare relevant not only for portfolio de-cisions within a particular asset class but also for portfolio decisions acrossdifferent asset classes Estimates of total incentives facing asset managers areavailable for some asset classes in addition to hedge funds
Probably the most important ldquoalternative assetrdquo other than hedge funds isprivate equity funds This investment class has very similar direct incentiveschemes as hedge funds In addition they have substantial indirect incentivesChung et al (2012) estimate that private equity fund managersrsquo indirect in-centives are approximately the same size as their direct incentives
Mutual funds are another asset class for which one can measure indirectincentives An important difference from hedge funds is that mutual fundsdo not charge incentive fees or carried interest so direct contractual incen-tives do not exist However the flow-performance relation facing mutual fundmanagers potentially leads to substantial indirect incentives To understandhow mutual funds and hedge funds compare in terms of the magnitude andimportance of indirect incentives we conduct similar calculations for mutualfunds as we do for hedge funds These calculations also require estimates of theflow-performance sensitivity as well as a valuation model of the mutual fundmanagerrsquos expected lifetime compensation
To estimate the flow-performance sensitivity in as comparable a manneras possible to our hedge fund estimates we construct a sample of mutualfunds from the CRSP Survivorship Bias Free Mutual Fund Database over thesame 1995 to 2010 period used in the hedge fund sample Because we focuson flows into actively managed funds we exclude index funds and exchange-traded funds (ETF) We also exclude fixed income funds international funds
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 903
and sector funds to facilitate comparison with prior literature We then applythe same filtering rules as we used with hedge funds that is we require theavailability of contractual information and at least 12 consecutive monthlyobservations within the sample period We also drop all fractional quarters
Our sample consists of 307079 fund-quarters associated with 11911 fundsThe average mutual fund in our sample contains $322 million in assets and iseight years old The management fee defined as the total expense ratio minusthe 12B-1 fee averages 11 of total net assets Further descriptive statisticsare provided in Table CII in Appendix C
We estimate the mutual fund analog of equation (1) using quarterly obser-vations from this mutual fund sample and report the results in Table CIIIin Appendix C19 In doing so we omit hedge fund-specific variables such asthe incentive fee that are not relevant for mutual funds and include style byquarter fixed effects for mutual fund styles listed in Table CII in Appendix CWe find that the same patterns hold for mutual funds as for hedge funds thecoefficients on past returns are all positive and statistically significant in mostcases and decline almost monotonically over time The sum of coefficients onpast returns is about 178 (based on Column (2) of Table CIII) implying thatflows are somewhat more responsive to performance in the mutual fund indus-try than in the hedge fund industry This pattern may be attributed to the factthat search costs and transaction costs are much lower for mutual funds thanfor hedge funds which leads to higher flow-performance sensitivity (HuangWei and Yan (2007))
We obtain the present value of mutual fund fees per dollar of AUM using thesame LWY framework that we use for hedge funds Within this framework amutual fund can be modeled as a fund in which leverage is not allowed and theincentive fee is zero Therefore the LWY model with no leverage (equivalentto the GIR model) with the incentive fee rate (k) set to zero characterizes thepresent value of fees for a mutual fund In addition since mutual funds do nothave an HWM provision the contractual growth rate of HWM (g) is set to zeroand the ratio between the investorrsquos wealth and HWM (w) is set to one Forcomparison with the hedge fund estimates we present estimates for the sameeight combinations of the parameters b α and δ + λ For all other parametersfund-specific values are used
Table VIII provides estimates of indirect incentives in mutual funds In con-trast to the hedge fund analyses we do not compare indirect incentives to direct
19 We estimate the same linear specification as we used with hedge funds to obtain compa-rable flow-performance sensitivity estimates Alternatively to incorporate the well-documentednonlinearity (convexity) in the flow-performance relation for mutual funds we estimate a sim-ple piecewise specification that partitions negative returns and positive returns The resultingflow-performance sensitivity estimate is about 26 larger when returns are positive than whatis implied from the linear specification which leads to 17 larger indirect incentives than thosereported in Table VIII When we take into account convexity effects by including quadratic laggedreturn terms in the regression the resulting estimate suggests that indirect incentives are about13 larger than those reported in Table VIII when evaluated at the 75th percentile of quarterlyreturns
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
904 The Journal of Finance Rcopy
Tab
leV
III
Ind
irec
tP
ayfo
rP
erfo
rman
cefo
rM
utu
alF
un
ds
Th
ista
ble
pres
ents
esti
mat
esof
indi
rect
pay
for
perf
orm
ance
(in
cen
tive
s)fo
rm
utu
alfu
nds
ata
quar
terl
yfr
equ
ency
usi
ng
the
LWY
mod
el(e
quiv
alen
tto
the
GIR
mod
elu
nde
rth
ese
tup
for
mu
tual
fun
ds)
Wit
hin
this
fram
ewor
ka
mu
tual
fun
dis
asp
ecia
lca
sein
wh
ich
leve
rage
isn
otal
low
edan
dth
ein
cen
tive
fee
(k)
isze
roM
utu
alfu
nds
hav
en
oH
WM
prov
isio
na
nd
ther
efor
eth
eco
ntr
actu
alH
WM
grow
thra
te(g
)is
set
toze
roan
dth
era
tio
betw
een
anin
vest
orrsquos
wea
lth
and
HW
Mle
vel
(w)
ison
eat
all
tim
es
Est
imat
esof
indi
rect
ince
nti
ves
are
pres
ente
dfo
rei
ght
com
bin
atio
ns
ofth
epa
ram
eter
sbα
an
dδ+λ
All
oth
erpa
ram
eter
valu
esar
efu
nd-
spec
ific
Rep
orte
dst
atis
tics
are
aver
ages
take
nov
eral
lfu
nd-
quar
ters
(307
079
)in
the
data
Flo
w-p
erfo
rman
cese
nsi
tivi
tyto
com
pute
the
indi
rect
effe
ctfr
omn
ewm
oney
isob
tain
edfr
omth
eco
effi
cien
tes
tim
ates
inC
olu
mn
(2)
ofTa
ble
CII
Iin
App
endi
xC
Row
s(4
)an
d(9
)(R
ows
(5)
and
(10)
)re
port
the
mag
nit
ude
ofin
dire
ctin
cen
tive
sfo
rm
utu
alfu
nds
inco
mpa
riso
nto
the
esti
mat
esfo
rh
edge
fun
dsfr
omth
eG
IRm
odel
(LW
Ym
odel
)
b=
068
5b
=0
8
α=
0α
=3
α=
0α
=3
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
δ+λ
=5
δ+λ
=10
(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)
Per
1ch
ange
infu
nd
valu
e($
mil
lion
s)(1
)In
dire
ctef
fect
from
new
mon
ey0
392
025
90
769
036
70
214
015
60
414
022
4(2
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
022
10
146
043
30
206
012
00
088
023
30
126
(3)
Tota
lin
dire
ctef
fect
061
30
405
120
20
573
033
40
243
064
70
351
(4)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
658
58
3
602
51
8
516
45
9
476
39
6
(5)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
539
55
7
465
42
0
343
34
7
419
35
9
Per
$1ch
ange
infu
nd
valu
e($
)(6
)In
dire
ctef
fect
from
new
mon
ey0
139
009
50
260
013
30
074
005
50
136
007
9(7
)In
dire
ctef
fect
from
chan
gein
valu
eof
exis
tin
gas
sets
007
80
053
014
60
075
004
10
031
007
60
044
(8)
Tota
lin
dire
ctef
fect
021
60
148
040
60
208
011
50
086
021
20
123
(9)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eG
IRm
odel
514
47
1
466
42
1
406
37
5
369
32
7
(10)
of
the
hed
gefu
nd
esti
mat
esfr
omth
eLW
Ym
odel
435
46
4
378
36
0
277
28
8
330
30
0
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 905
incentives Direct incentives are small in the mutual fund industry because mu-tual funds do not charge incentive fees and mutual fund manager ownership istypically very small (eg Ma and Tang (2014)) In terms of indirect incentivesTable VIII shows that for an incremental one-percentage-point return to thefundrsquos investors the manager of an average-sized mutual fund ($322 million)receives as high as $1202000 in expected future compensation which corre-sponds to 47 of what the LWY model suggests an average hedge fund managerreceives under the same parameter assumptions and 60 of what the GIRmodel implies for hedge funds Depending on the parameter choices the rela-tive magnitude of the indirect effect in mutual funds ranges between 40 and66 of the hedge fund counterpart within the GIR framework (Row (4)) and be-tween 34 and 56 within the LWY framework (Row (5)) Expressed as a frac-tion of an incremental dollar return to the fund mutual fund managers receive86 to 406 cents which corresponds to 33 to 51 of the size of indirect incen-tives facing hedge fund managers within the GIR framework (Row (9)) and 28to 46 within the LWY framework (Row (10)) Indirect incentives are substan-tially smaller in mutual funds than in hedge funds but nonetheless constitutean important component of the incentives motivating mutual fund managers
C Implications for Understanding Contracting More Broadly
Ever since Fama (1980) and Holmstrom (1982) it has been recognized thatmarket-based incentives are an important element of corporate governanceand also motivate many workers who are not top managers Yet while therehas been some work estimating their importance for CEOs of large industrialcorporations it is difficult to evaluate their magnitude quantitatively20 Theestimates provided here for hedge fund managers are some of the first esti-mates of the importance of indirect incentives for managers other than CEOsof large corporations The cross-sectional pattern of the estimates is consis-tent with a Bayesian learning process by which the market provides theseincentives
An open question is the extent to which our estimates compare to market-based incentives of more typical managers While most managers do not havethe upside potential of hedge fund managers they also do not have directincentive schemes that are nearly as powerful Consequently relative to directincentive pay indirect market-based incentives could potentially be equally oreven more important to most workers as they are for hedge fund managers
VI Conclusion
Managersrsquo incentives to perform well come not only through direct pay-forperformance plans but also through the managerial labor market The market
20 See Boschen and Smith (1995) and Taylor (2013) for evidence of the importance of indirectincentives for CEOs and Hermalin and Weisbach (2014) for a general discussion of indirect incen-tives and their measurement
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
906 The Journal of Finance Rcopy
draws inferences about managersrsquo abilities from their observed performanceand rewards or penalizes them accordingly providing an additional channelthrough which managersrsquo performance can affect their welfare The moneymanagement industry is one place where the managerial labor market providessubstantial incentives since investors can observe managersrsquo performance andreallocate capital easily between alternative investments This paper estimatesthe magnitude of this effect for hedge fund managers
Our estimates indicate that indirect incentives are particularly large forhedge fund managers On average for an incremental dollar returned to in-vestors managersrsquo expected lifetime income increases by at least 39 cents 23cents of which comes from indirect incentives arising from managersrsquo abilityto earn fees on the increased AUM that follows incremental performance Thelarge indirect incentives for hedge fund managers come from the fact that in-flows to hedge funds are highly sensitive to performance and the fee structurein hedge funds is such that managers expect to receive a large fraction of eachdollar invested in the fund as compensation over time
The existence of substantial indirect incentives facing hedge fund managershas a number of implications Hedge funds appear to be organized to provideoptimal incentives to their managers Yet direct incentives do not appear toadjust for differences in indirect incentives across funds or over a fundrsquos lifecycle Investors who care about fund managersrsquo incentives should care abouttheir total incentives including both direct and indirect components Bothprivate equity funds and mutual funds also appear to have substantial indirectincentives although not as large as those facing hedge fund managers Finallyasset management is but one industry in which it is possible to measure indirectincentives facing managers Undoubtedly such incentives are also importantfor many other types of jobs Whether indirect incentives are of comparableimportance outside of the money management industry would be an excellenttopic for future research
Initial submission July 30 2013 Final version received May 23 2015Editor Kenneth Singleton
Appendix A Calculating Direct Pay for Performance
Our proxy for direct pay for performance is given by total delta as definedin Agarwal Daniel and Naik (2009) Total delta at a point in time is definedas the expected dollar change in the managerrsquos wealth for a one-percentage-point increase in the fundrsquos return over the following year The total delta canbe decomposed into two parts the portion coming from investorsrsquo assetsmdashthemanagerrsquos option deltamdashand the portion coming from the managerrsquos owner-ship stake in the fund The managerrsquos option delta is in turn the sum of thedeltas associated with the different investors in the fund because of the factthat different investors in the fund face different spot prices (S) and exerciseprices (X) depending on the timing of their entrance into the fund Following
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 907
Agarwal Daniel and Naik (2009) for each fund-investor-quarter we computethe individual optionrsquos delta using the Black-Sholes model for a one-quartermaturity European call option as follows
Individual Option Delta = N(Z) times S times 001 times k (A1)
where Z = ln(SX) + T(r+σ 22)σT05 S is the investor-specific spot price(ie the market value of the investorrsquos assets) k is the contractual incentive feerate and X is the investor-specific exercise price Given the structure of hedgefund contracts X is the HWM level for the investorrsquos assets (ie the historichigh that the investorrsquos asset has reached since her investment in the fund) ifthe fund has an HWM provision and simply the market value of the investorrsquosassets if the fund has no such provision In either case X can be increased by ahurdle rate if applicable Because incentive fees are paid out at the end of eachquarter SX can never be greater than one Therefore SX measures whetherthe option is at the money and if not the degree to which the option is out ofthe money The variable T is the time to maturity of the option (one quarterin this case) r is the natural logarithm of one plus the risk-free rate which ismeasured as the three-month LIBOR rate that is in effect at the beginning ofeach quarter σ is the standard deviation of monthly (net-of-fee) returns overthe prior 12-month period multiplied by the square root of three to arrive ata quarterly figure and N() is the standard normal cumulative distributionfunction
A complication in computing individual option deltas is that various in-vestorsrsquo assets are pooled together for management so they earn the samerate of gross return but different investors will in general have different spotprices (S) and exercise prices (X) depending on when they entered the fundUnfortunately the exact times and prices at which each investor entered eachfund are not publicly available To compute the investor-specific spot price (S)and exercise price (X) for each investor we make the following assumptions
1 The first investor enters the fund at the end of quarter 0 (beginning ofquarter 1) There is no capital investment by the manager at inceptionTherefore the entire assets at inception come from a single investor
2 All cash flows including fee payments investorsrsquo capital allocation andthe managerrsquos reinvestment take place once a quarter at the end of eachcalendar quarter
3 The exercise price (X) for each option is reset at the end of each quarterand applies to the following quarter
4 All new capital inflows come from a single new investor5 When capital outflows occur we adopt the FIFO rule to decide which
investorrsquos money leaves the fund In particular the asset value of thefirst investor is reduced by the magnitude of outflow If the absolutemagnitude of outflow exceeds the first investorrsquos net asset value thenthe first investor is considered as liquidating her stake in the fund andthe balance of outflow is deducted from the second investorrsquos assets andso on
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
908 The Journal of Finance Rcopy
6 The hurdle rate is LIBOR if the fund has an HWM provision and 0otherwise
7 Managers reinvest all of their incentive fees into the fund
An algorithm for estimation is as follows
1 First we solve the following recursive problem iteratively to back outgross returns (gross) using observable information on net-of-fee returns(net) assets under management (AUM) and the parameters of the com-pensation contract (kc)
nett =sumi
Sitminus1 times (1+grosst) minus ifeeit minus mfeeit+MStminus1 times (1+grosst)
AUMtminus1minus 1
(A2)
where ifeeit = Max[(Sitminus1times(1+grosst) minus Xitminus1) 0]timesk mfeeit = Sitminus1timescand MSt denotes the market value of the managerrsquos stake in the fundWe start the algorithm with the following initial values
S10 = X10 = AUM0
MS0 = 0
(A3)
2 We update the market value of the managerrsquos stake as follows
MSt = MStminus1 times (1 + grosst) +sum
i
i f eeit (A4)
3 The new spot price and exercise price of investor i are updated recursivelyas follows
Sit = Sitminus1 times (1 + grosst) minus i f eeit minus mf eeit (A5a)
Xit =
Max[Sit Xitminus1] times (1 + LIBOR) if with HWM provisionSit if without HWM provision (A5b)
4 The net flow into the fund is defined as the difference between the re-ported value of quarter-end AUM and the current market value of allexisting investorsrsquo assets and the managerrsquos assets
Flowt = AUMt minus(sum
i
Sit + MSt
) (A6)
If (A6) gt 0 then we assume that a new investor enters the fund withthe beginning spot price and exercise price equal to Flowt If (A6) lt 0then we apply the FIFO rule as described above
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 909
5 Using S and X for each investor and equation (A1) we compute thedelta from each investorrsquos assets and then sum them up to computethe managerrsquos option delta The total delta of the fund is the sum ofthe managerrsquos option delta and the delta from the managerrsquos stake whichis equal to 001MSt because the manager retains all returns from in-vesting his own assets
Appendix B Calculating Indirect Pay for Performance
We employ four models to compute the present value of the total futurefees (both management and incentive fees) per dollar in the fund Each of thesemodels provides this value as a function of the ratio of the asset value to itsHWM that is the stake-to-strike (SX) ratio of the assets which we denote byw This subsection describes each of these models in turn discusses our choicesof model parameters (a summary of parameter choices is provided in Table CI)and describes the calculations of indirect incentives
B1 Baseline GIR Model
Goetzmann Ingersoll and Ross (2003) provide a closed-form solution for thevalue of total fees as follows
f (w) = 1c + δ + λminus α
[c + (δ + λminus α)k + [η(1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wγminus1
minusbγminusη(δ + λminus α)k + [γ (1 + k) minus 1]cb1minusη
γ (1 + k) minus 1 minus bγminusη[η(1 + k) minus 1]wηminus1
]
(B1)
where γ and η are the larger and smaller roots of the following characteristicquadratic equation
(γ
η
)equiv
12σ
2 + c minus r minus α minus cprime + g plusmnradic( 1
2σ2 + c minus r minus α minus cprime + g
)2 + 2σ 2 (r + cprime minus g + δ + λ)
σ 2 (B2)
There are 10 parameters in the valuation equation (B1) c and k are thecontractual management fee and incentive fee rate respectively cprime is the ac-counting choice of costs and fees allocated to reducing HWM σ is the volatilityof fund returns g is the contractual growth rate in the HWM level (ie hurdlerate) which is usually zero or the risk-free rate α is the risk-adjusted expectedgross-of-fee return on the fundrsquos assets reflecting manager skill δ + λ is thetotal withdrawal rate which is the sum of the regular payout rate to investors(δ) and the exogenous liquidation probability of the fund (λ) b is the lowestacceptable fraction of the HWM below which the investor loses confidence inthe fund and liquidates all of his position and r is the risk-free interest rateThe parameter b captures a performance-triggered liquidation boundary Forexample with b = 08 an investor liquidates his entire stake if it falls in valueby 20 from its HWM
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
910 The Journal of Finance Rcopy
We estimate the parameters of equation (B1) from observable fund-leveldata whenever possible In particular for c and k we use an individual fundrsquoscontractual information available from TASS We compute σ as the squareroot of three times the standard deviation of monthly returns over the priorcalendar year The value of w for each existing investor is computed based onS and X obtained from the estimation procedure described in Appendix A Forr we use the three-month LIBOR at the beginning of each quarter
For the rest of the parameters that are not observable or reasonably es-timable we adopt the quarterly equivalent of the baseline parameter valuesused in GIR or LWY α = 0 and 3 δ + λ = 5 and 10 b = 0685 and 08g = r if a fund has a HWM provision and zero otherwise and cprime = 021 Theseparameter choices are summarized in Table CI in Appendix C
B2 GIR Model Augmented for Performance-Based Fund Flows
One of the limitations of the base-case GIR model is that it does not allow forperformance-based fund flows One way to address this issue is to assume asan approximation that the flow-performance relationship is linear in logs andthat flows are contemporaneous with returns Without performance-inducedflows the log AUM of the fund would evolve as a martingale so st+1 = st +et+1 With performance-based flows st+1 = st + get+1 where g gt 1 captures theflow effect The log AUM still follows a martingale Given these assumptionsthe GIR model still applies but with a higher variance than in the case of noflows
We implement this approach to incorporating fund flows in the follow-ing manner First we estimate the relation between log(AUMtAUMtminus1) andlog(1 + Returnt) controlling for all the other variables used in Table II exceptthe lagged return terms When we estimate this equation the coefficient onthe logged contemporaneous return equals 1247 Then at the fund-quarterlevel we multiply the GIR volatility σ by this coefficient to provide a volatilityparameter for the extended GIR model that is likely to reflect performance-based flows Empirically the mean quarterly volatility is boosted from 533to 665 One complication in using the new volatility parameter is the way itinteracts with the liquidation boundary parameter b Because the increase involatility is associated with flows and not returns per se the greater volatilityshould not result in a greater likelihood of performance-based liquidation Toensure this does not happen we decrease b such that under the new meanvolatility performance-based liquidation is as likely as before For examplewith b = 08 (0685) an investor withdraws all assets following a minus 200(315) return which is 375 (591) standard deviations away from one under
21 In the GIR model cprime is determined by an unobservable accounting choice Its value is notexplicitly discussed but r + cprime minus g is assumed to be 5 which implies a choice of cprime = 5 becauser always equals g in GIR In contrast the LWY model does not have this parameter and thereforeeffectively assumes it is equal to zero Since cprime cannot be reasonably estimated from the data and apositive value for it serves to increase indirect incentives substantially we follow LWY and assumethat it equals zero throughout the paper
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 911
the volatility parameter without flows Now under the new volatility parame-ter a 375 (591) standard deviation drop translates into a minus 249 (393)return or equivalently b = 0751 (0607) In summary relative to the base GIRmodel the augmented model multiplies all fund-quarter-level volatility esti-mates by 1247 and adjusts b as described above
B3 Baseline LWY Model
A drawback of the GIR model is that it does not allow the manager to alterhis or her portfolio endogenously However in practice hedge fund managershave incentives to change their investment strategies and portfolio allocationdynamically to maximize their value function One way they potentially adjusttheir portfolios is through the use of leverage dynamically trading off thebenefit and the downside risk of leverage by allocating the fundsrsquo AUM betweenthe alpha-generating strategy and the risk-free asset To address this issue weemploy the model provided by Lan Wang and Yang (2013) which is a dynamicframework to value a hedge fund managerrsquos compensation contract under theendogenous leverage choice of the manager22
In the baseline LWY setting the managerrsquos value function f(w) solves thefollowing ODE
(β minus g + δ + λ) f (w) = cw + [π (w)αprime + r minus g minus c]w f prime (w)
+ 12π (w)2σ prime2w2 f primeprime (w)
(B3)
subject to the boundary conditions
f (b) = 0 (B4)
f (1) = (k + 1) f prime (1) minus k (B5)
where αprime and σ prime represent unlevered alpha and volatility respectivelyThe optimal leverage policy π (w) is dynamically determined as follows
π (w) =
min
αprime
σ prime2ψ(w) π ψ (w) gt 0
π ψ (w) le 0 (B6)
22 There are other hedge fund valuation models that incorporate endogenous portfolio alloca-tion by hedge fund managers For example Panageas and Westerfield (2009) provide closed-formsolutions for endogenously determined hedge fund leverage and valuation in a setting with noperformance-induced liquidation boundary (ie b = 0) and no management fees (ie c = 0) Drech-sler (2014) also solves for the managerrsquos optimal leverage choice and derives closed-form solutionsalbeit by counterfactually assuming that management fees are a constant fraction of the HWMnot the AUM Overall Drechsler (2014) reaches similar results to LWY when incorporating con-siderations such as performance-triggered liquidation risk management fees and the managerrsquosrestart option We use the LWY model because of its generality as it embeds as special cases bothPanageas and Westerfield (2009) and GIR (with leverage exogenously fixed at zero)
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
912 The Journal of Finance Rcopy
where ψ(w) is a risk-neutral managerrsquos endogenous risk attitude Risk attitudeψ(w) is in turn defined by
ψ (w) equiv minus f primeprime (w)f prime (w)
(B7)
Relative to the GIR model the baseline LWY model requires calibrations ofthree additional parameters the unlevered alpha (αprime) the unlevered volatility(σ prime) and the managerrsquos discount rate (β) We equate β to r by assuming thatthe manager and the investor use the same discount rate Following LWY weadopt the average leverage of 213 reported by Ang Gorovyy and van Inwegen(2011) The unlevered alphas (αprime) implied by levered alphas of 0 and 3 are0 and 14 respectively Likewise the unlevered volatility (σ prime) is obtainedfor each fund-quarter by dividing the estimated quarterly volatility σ in thedata by 21323 Parameter choices are summarized in Table CI In generalthe ODE in equation (B3) does not have an analytical solution and must beapproximated numerically
B4 LWY Model Augmented for Performance-Based Inflows
LWY provide an extension of their baseline model that includes futureperformance-based fund flows LWY assume that new money arrives over in-cremental time interval (t t + t) and show that dIt evolves as follows
dIt = i[dHt minus (g minus δ) Htdt
] (B8)
where I gt 0 denotes the sensitivity of dIt with respect to the fundrsquos instanta-neous (gross) profits
Now the value of total fees f(w) satisfies ODE (B3) subject to the followingboundary conditions
f (b) = 0 (B9)
f (1) = (k + 1) f prime (1) minus k1 + i
(B10)
We choose i = 08 following LWY In general a larger value for i leads toa larger π (w) and therefore a larger f(w) because the manager becomes lessrisk-averse when there are more rewards at the upside Our implementation ofthe augmented LWY model mirrors that of the baseline LWY model describedabove
23 The unlevered volatility (σ prime) is floored and capped at 14 and 42 respectively because theODE (B3) cannot be solved when σ prime takes an excessively small or large value
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 913
B5 Calculating Indirect Incentives
Indirect incentives have two components future fees earned from the in-cremental inflows of new investment that follow incremental performance andfuture fees earned on the increase in the value of existing investorsrsquo stakes thatfollows incremental performance The latter component occurs both because thevalue of existing investorsrsquo stakes is higher and also because it moves closer tothe applicable HWM
For the first component inflows of new investment the present value of totalfees for each dollar of new money coming into the fund f(w)new can be computedusing one of the four valuation models discussed above with w determined asfollows
wnew =
1 if the fund has no hurdle rate1
1+r if the fund has a hurdle rate(B11)
Indirect pay for performance coming from new money flows per 1 incremen-tal return is calculated as f(w)new times the sum of the regression coefficients onlagged returns (with appropriate age or strategy interactions) given in Column(2) of the appropriate Panel of Table II multiplied by 1 of beginning-of-yearAUM
To compute indirect pay for performance from changes in the value of existinginvestorsprime stakes we proceed as follows First we compute the f(w) fraction foreach fund-quarter for each existing investor f (wi)base
old assuming that the fundearns an baseline return equal to the mean return earned by all funds of thesame age and investment strategy To obtain investor-specific wi under thebaseline return we take each investorrsquos wi at the beginning of the quarter(calculated as described in Appendix A) and multiply by (1 + baseline return)Then if the result is less than one and the fund has an HWM provision wi isset to the result divided by (1 + r) If the result is greater than one and the fundhas a HWM set wi to 1(1 + r)24 Finally if the fund does not have an HWMset wi to one We adjust wi in this way because if the result is greater than onethen the option becomes in the money incentive fees are paid and the strikeresets We sum these individual investorsrsquo f(w) fractions over all investors inthe fund as follows
f (w)baseold =
sumi
xi f (wi)baseold where xi = Sisum
iSi (B12)
We next rerun these calculations for existing investors assuming that thefund earns an additional one-percentage-point return in addition to the base-line return ( f (w)base+1
old ) Then we estimate the impact of an incremental 1return on the managerrsquos future pay due to the increase in the asset values of
24 As in Appendix A we assume that an fund has a hurdle rate whenever it has an HWMprovision
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
914 The Journal of Finance Rcopy
existing investors as the difference between f (w)base+1old times AUM times (1+baseline
return + 1) and f (w)baseold times AUM times (1 + baseline return)
Finally indirect incentives per incremental one-dollar increase in fund re-turns are calculated for each fund-quarter by dividing indirect incentives perone-percentage-point increase in returns by 1 of beginning-of-quarter AUM
Appendix C Tables
Table CISummary of Parameter Choices
This table summarizes the parameters for the four models used to calculate indirect incentives
Parameter Baseline GIR Augmented GIR Baseline LWY Augmented LWY
c Fund specificannual rate4
Fund specific annualrate4
Fund specific annualrate4
Fund specific annualrate4
k Fund specific Fund specific Fund specific Fund specificσ (σ prime) Fund-quarter
specificquarterlyvolatility =standarddeviation ofmonthlyreturns over theprior 12 monthperiod times radic
3
Fund-quarter specificquarterly volatilitytimes 1247
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
Fund-quarter specificcorrespondingunlevered volatility(σ prime) = quarterlyvolatility213
α (αprime) Quarterlyequivalent of0 3
Quarterly equivalentof 0 3
Quarterly equivalentof 0 3correspondingunlevered alpha (αprime)= 0 14
Quarterly equivalent of0 3 correspondingunlevered alpha (αprime) =0 14
δ+λ Quarterlyequivalent of5 10
Quarterly equivalentof 5 10
Quarterly equivalentof 5 10
Quarterly equivalent of5 10
b 0685 08 0607 0751 0685 08 0685 08r 3-Month LIBOR 3-Month LIBOR 3-Month LIBOR 3-Month LIBORg = r if HWM = r if HWM Equated to r Equated to r
= 0 if no HWM = 0 if no HWMβ na na Equated to r Equated to rcprime 0 0 na naw(=SX) Fund-quarter-
investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
Fund-quarter-investorspecific
i na na na 08
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 915
Table CIISummary Statistics for the Mutual Fund Sample
This table presents descriptive statistics for the sample of 307079 fund-quarter observationsassociated with 11911 mutual funds during the 1995 to 2010 sample period Data are obtainedfrom the Center for Research in Security Prices (CRSP) Survivorship Bias Free Mutual FundDatabase Only actively managed domestic equity funds are considered excluding index fundsexchange-traded funds exchange-traded notes international funds fixed income funds and sectorfunds Quarterly flow Quarterly returns Age and Volatility are constructed in the same way asthe corresponding variables for hedge funds Management fees are the expense ratio minus 12b1fees as the percentage of total assets measured at the end of each quarter All variables exceptthe style indicator are reported at the fund-quarter level Quarterly flow Quarterly returns andVolatility are winsorized at the 1st and 99th percentiles
Variable N Mean 25th Pctl Median 75th Pctl SD
Quarterly flow () 307079 505 minus495 minus055 666 2533Quarterly returns () 307079 159 minus406 252 800 1034AUM ($ million) 307079 3222 66 384 1847 9362Age (quarters) 307079 3128 1144 2138 3735 3613Volatility () 307079 834 522 760 1053 406Management fees ( annualized) 307079 108 086 105 125 082Style
Large-cap (EDCL) 11911 0002Mid-cap (EDCM) 11911 0108Small-cap (EDCS) 11911 0171Micro-cap (EDCI) 11911 0007Growth amp Income (EDYB) 11911 0230Growth (EDYG) 11911 0438Income (EDYI) 11911 0030Hedged (EDYH) 11911 0014Dedicated short bias funds (EDYS) 11911 0001
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
916 The Journal of Finance Rcopy
Table CIIIEstimates of Flow-Performance Sensitivity for Mutual Funds
This table presents the OLS coefficient estimates of equation (1) and corresponding standard errors(in parentheses) for the sample of mutual funds over the period 1995 to 2010 The dependentvariable is Quarterly flow as defined by equation (2) Style fixed effects quarter fixed effectsand quarter-by-style fixed effects are included in all models Standard errors are double clusteredby fund and year and indicate statistical significance at the 1 5 and 10 levelrespectively
(1) (2)
Dependent var = Quarterly flow(t) Coef (SE) Coef (SE)
Quarterly returns(tminus1) 0647 (0060) 0539 (0057)Quarterly returns(tminus2) 0405 (0056) 0284 (0051)Quarterly returns(tminus3) 0252 (0046) 0176 (0042)Quarterly returns(tminus4) 0216 (0045) 0171 (0038)Quarterly returns(tminus5) 0170 (0040) 0125 (0034)Quarterly returns(tminus6) 0135 (0053) 0106 (0044)Quarterly returns(tminus7) 0100 (0044) 0083 (0036)Quarterly returns(tminus8) 0060 (0044) 0049 (0037)Quarterly returns(tminus9) 0086 (0041) 0078 (0034)Quarterly returns(tminus10) 0114 (0040) 0104 (0031)Quarterly returns(tminus11) 0075 (0042) 0063 (0035)Quarterly flow(tminus1) 0253 (0009)Log(Age + 1) minus0023 (0002)Log(AUM(tminus1)+1) minus0010 (0001)Volatility minus0024 (0059)Management fees minus0143 (0329)Indicators for missing lagged returns Yes YesFixed effects
Style Yes YesQuarter Yes YesStyle times Quarter Yes Yes
Number of observations 307079 307079Adjusted R2 0123 0192F-test for joint significance of F-stat (p-value) F-stat (p-value)All lagged returns 2029 (0000) 1853 (0000)
REFERENCES
Agarwal Vikas Naveen Daniel and Narayan Naik 2003 Flows performance and managerialincentives in the hedge fund industry Working paper Georgia State University
Agarwal Vikas Naveen Daniel and Narayan Naik 2009 The role of managerial incentives anddiscretion in hedge fund performance Journal of Finance 44 2221ndash2256
Agarwal Vikas and Sugata Ray 2012 Determinants and implications of fee changes in the hedgefund industry Working paper Georgia State University
Ang Andrew Sergiy Gorovyy and Greg van Inwegen 2011 Hedge fund leverage Journal ofFinancial Economics 102 102ndash126
Aragon George 2007 Share restrictions and asset pricing Evidence from the hedge fund industryJournal of Financial Economics 83 33ndash58
Aragon George and Vikram Nanda 2012 Tournament behavior in hedge funds High-watermarks fund liquidation and managerial stake Review of Financial Studies 25 937ndash974
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
Indirect Incentives of Hedge Fund Managers 917
Baquero Guillermo and Marno Verbeek 2009 A portrait of hedge fund investors Flows perfor-mance and smart money Working paper Erasmus University
Barclay Michael J Neil D Pearson and Michael S Weisbach 1998 Open-end mutual funds andcapital gains taxes Journal of Financial Economics 49 3ndash43
Berk Jonathan and Richard Green 2004 Mutual fund flows and performance in rational marketsJournal of Political Economy 112 1269ndash1295
Bollen Nicolas PB 2007 Mutual fund attributes and investor behavior Journal of Financial andQuantitative Analysis 42 683ndash708
Bollen Nicolas PB and Veronika K Pool 2009 Do hedge fund managers misreport returnsEvidence from the pooled distribution Journal of Finance 64 2257ndash2288
Boschen John F and Kimberly J Smith 1995 You can pay me later The dynamic response ofexecutive compensation to firm performance Journal of Business 68 577ndash608
Brown Keith C W Van Harlow and Laura T Starks 1996 Of tournaments and temptations Ananalysis of managerial incentives in the mutual fund industry Journal of Finance 51 85ndash10
Chevalier Judith and Glenn Ellison 1999 Career concerns of mutual fund managers QuarterlyJournal of Economics 114 389ndash432
Chung Ji-Woong Berk A Sensoy Lea H Stern and Michael S Weisbach 2012 Pay for perfor-mance from future fund flows The case of private equity Review of Financial Studies 253259ndash3304
Del Guercio Diane and Paula A Tkac 2002 The determinants of the flow of funds of managedportfolios Mutual funds vs pension funds Journal of Financial and Quantitative Analysis37 523ndash557
Deuskar Prachi Zhi Jay Wang Youchang Wu and Quoc H Nguyen 2012 The dynamics of hedgefund fees Working paper University of Oregon
Ding Bill Mila Getmansky Bing Liang and Russ Wermers 2009 Share restrictions and investorflows in the hedge fund industry Working paper University of Massachusetts
Drechsler Itamar 2014 Risk choice under high-water marks Review of Financial Studies 272052ndash2096
Fama Eugene F 1980 Agency problems and the theory of the firm Journal of Political Economy88 288ndash307
Fung William David A Hsieh Narayan Y Naik and Tarun Ramadorai 2008 Hedge fundsPerformance risk and capital formation Journal of Finance 63 1777ndash1803
Getmansky Mila 2012 The life cycle of hedge funds Fund flows size competition and perfor-mance Quarterly Journal of Finance 2 57ndash109
Gibbons Robert and Kevin J Murphy 1992 Optimal incentive contracts in the presence of careerconcerns Theory and evidence Journal of Political Economy 100 468ndash505
Goetzmann William N Jonathan E Ingersoll Jr and Stephen A Ross 2003 High water marksand hedge fund management contracts Journal of Finance 43 1685ndash1717
Gompers Paul and Josh Lerner 1999 An analysis of compensation in the US venture capitalpartnership Journal of Financial Economics 51 3ndash44
Guasoni Paolo and Jan Oblόj 2013 The incentives of hedge fund fees and high-water marksMathematical Finance doi 101111mafi12057
Hermalin Benjamin E and Michael S Weisbach 2014 Understanding corporate governancethrough learning models of managerial competence NBER Working paper 20028
Holmstrom Bengt 1982 Managerial incentive problems A dynamic perspective in Essays inHonor of Lars Wahlbeck (Swedish School of Economics Helsinki)
Huang Jennifer Kelsey D Wei and Hong Yan 2007 Participation costs and the sensitivity offund flows to past performance Journal of Finance 62 1273ndash1311
Ippolito Richard A 1992 Consumer reaction to measures of poor quality Evidence from themutual fund industry Journal of Law and Economics 35 45ndash70
Kaplan Steve N and Antoinette Schoar 2005 Private equity performance Returns persistenceand capital flows Journal of Finance 60 1791ndash1823
Kaplan Steven N and Joshua Rauh 2010 Wall Street and Main Street What contributes to thehighest incomes Review of Financial Studies 23 1004ndash1050
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University
918 The Journal of Finance Rcopy
Lan Yingcong Neng Wang and Jinqiang Yang 2013 The economics of hedge funds Journal ofFinancial Economics 110 300ndash323
Liang Bing 2001 Hedge fund performance 1990ndash1999 Financial Analysts Journal 57 11ndash18Lim Jongha Berk A Sensoy and Michael S Weisbach 2013 Indirect incentives of hedge fund
managers NBER Working paper 18903Ma Linlin and Yuehua Tang 2014 Portfolio manager ownership and mutual fund performance
Working paper Northeastern UniversityMetric Andrew and Ayako Yasuda 2010 The economics of private equity funds Review of Finan-
cial Studies 23 2303ndash2341Naik Narayan Y Tarun Ramadorai and Maria Stromqvist 2007 Capacity constraints and hedge
fund strategy returns European Financial Management 13 239ndash256Panageas Stavros and Mark M Westerfield 2009 High-water marks High risk appetites Convex
compensation long horizons and portfolio choice Journal of Finance 64 1ndash36Patton Andrew J Tarun Ramadorai and Michael Streatfield 2015 Change you can believe in
Hedge fund data revisions Journal of Finance 70 963ndash999Ramadorai Tarun 2013 Capacity constraints investor information and hedge fund returns
Journal of Financial Economics 107 401ndash416Ramadorai Tarun and Michael Streatfield 2011 Money for nothing Understanding variation in
reported hedge fund fees Working paper University of OxfordRobinson David T and Berk A Sensoy 2013 Do private equity fund managers earn their fees
Compensation ownership and cash flow performance Review of Financial Studies 26 2760ndash2797
Sensoy Berk A 2009 Performance evaluation and self-designated benchmark indexes in themutual fund industry Journal of Financial Economics 92 25ndash39
Sirri Eric and Peter Tufano 1998 Costly search and mutual fund flows Journal of Finance 531589ndash1622
Stein Jeremy C 1989 Efficient capital markets inefficient firms A model of myopic corporatebehavior Quarterly Journal of Economics 104 655ndash669
Swenson David F 2000 Pioneering portfolio management An unconventional approach to insti-tutional investment (Simon amp Schuster New York NY)
Taylor Lucian A 2013 CEO wage dynamics Estimates from a learning model Journal of Finan-cial Economics 108 79ndash98
Teo Melvyn 2009 Does size matter in the hedge fund industry Working paper Singapore Man-agement University