Message in a Bottle - TIFF

1

Copyright © 2000 by The Investment Fund for Foundations. All rights reserved. This report may not be reproduced or distributed, in whole or part, by any means, without written permission from TIFF.

MESSAGE IN A BOTTLEA Review of Investment Premises and Policies

Enhancing the investment returns

of non-profit organizations

Copyright © 2000by The Investment Fund for Foundations. All rights reserved. This report may not be reproduced or distributed, in whole or part, by any means, without written permission from TIFF.

2

Fortuitous Find

Magnum-Sized. Believe it or not, some faithful readersof TIFF’s voluminous reports voiced complaints aboutour last batch of quarterly commentaries, which werelarded with an abnormally large number of baseballanecdotes. Did readers complain that we injected toomuch humor into what should have been utterly seriousreports? Not at all: they complained that we had failedyet again to diversify our anecdotal reservoir with talesgleaned from other sports. We’re doing our best todiversify our holdings to a greater extent (e.g., golf toohas a rich tradition of anecdotal humor, although thereis not much interest in the game within TIFF and somemembers of the staff are stridently anti-golf). Butreaders may not enjoy the fruits of our diversificationefforts for some time, and until such fruits ripen, wewill do our best to avoid invoking baseball anecdotesto bring home certain points. Indeed, there aresurprisingly few baseball allusions in this document,and one of them is oblique at best: the star of the moviefrom which we’ve borrowed the title to the very longpiece that follows was also the star of a memorableflick about baseball (Kevin Costner, a/k/a Crash Davisin Bull Durham). The piece in question represents PartI of a two-part message that one peripatetic member ofthe TIFF staff found in a beached bottle. Like Mr.Costner’s female pursuer in the recent hit film Messagein a Bottle (Theresa Osborne, played by Robin WrightPenn), the TIFF staffer in question knows not thisdocument’s true origins — at least, not yet. But we arepublishing it nonetheless because it is a comprehensivereview of the premises underlying the investmentprogram pursued by what would appear to be a largephilanthropy. Indeed, “comprehensive” is a vastunderstatement because the “message” that followscomprises a very lengthy critique of an immenselypopular approach to long-term asset allocation knownas mean-variance analysis (“MVA”), a/k/a efficientfrontier analysis. Evidently, this particular organizationhas been relying heavily on MVA for some time, andthe following document was crafted primarily to inducecertain trustees to rely less heavily on MVA and moreheavily on alternate approaches to maintaining andenhancing endowment purchasing power.

Looking Ahead

Two-Part Message. As previously noted, the “messagein a bottle” that we are publishing comprises two parts,with the following table of contents.

Preface

I. PremisesA. Foundation Goals and Competencies

• Targeted Life• Liquidity Needs• Investment Time Horizon• Investment Objectives• Other Attributes• Behavioral Pitfalls

B. Markets• No Free Lunch• Risks• Hedging Costs

C. Asset Allocation Approaches• Overview• Mimicry Decried• “Efficient” Portfolios Decried

D. Manager Selection• Manager Selection Criteria• Expected Alphas

II. Investment PoliciesA. Return Objectives

• Primary Return Objective• Secondary Return Objective• Tertiary Return Objective• Underpinnings

B. Liquidity• Spending Policy• Cash Flow Requirements• Privately Traded Assets

C. Policy Portfolio• Purposes• Origins• Periodic Reviews• Conscious Offsets• Segment Return Objectives and Benchmarks• Asset Substitution• Asset-Specific Benchmarks

D. Implementation• Staff• Investment Committee• Rebalancing

MESSAGE IN A BOTTLE

This monograph was originally published in two parts.Part I appeared in TIFF’s Commentary dated March31, 1999, and Part II appeared in the June 30, 1999,Commentary.

3


First Principles

Food for Thought. Not all of TIFF’s directors andofficers agree fully with the “message in a bottle”reproduced below. But even if certain directors andofficers disagree with certain aspects of this document,they agree that it provides considerable food for thought,and they encourage readers to consider the extent towhich any endowed institutions with which they areassociated have formulated similarly comprehensivestatements of the premises underlying their investmentprograms. Experience suggests that it is by no meanseasy to fashion such a statement, and the difficulty ofthe task increases geometrically as the number oftrustees engaged in the discussion increases. But thevalue of clear “first principles” tends to be directlyproportionate to the effort expended in formulatingthem, and as part of its mission of helping allphilanthropies (members or otherwise) to make moreproductive use of their endowments, TIFF implores alltrustee groups to develop their own statement of “firstprinciples.” These statements need not be as long asthe one that follows, of course, but they should be atleast as clear as what follows — and perhaps even freerof apparent logical defects.

Odd-Shaped Cookie. At the risk of insulting readerswho know intuitively why we are labeling the policyreview that follows mere “food for thought,” we feelcompelled to emphasize that it is geared to a foundationwhose idiosyncratic attributes may differ in materialrespects from those of other endowed institutions withwhich readers are affiliated. Differently put, readerstempted to apply what follows in “cookie cutter”fashion to the investment challenges facing their ownorganizations should be aware that it is designed toproduce an odd-shaped cookie: the foundation inquestion has a relatively large asset base, does notnecessarily seek perpetual life status, and displaysother idiosyncracies that differentiate it from the typicalfoundation (if there is such a thing).

Ringing Endorsement

Part-Timers Not Wanted. To your editor’s way ofthinking, arguably the most important paragraph of thedocument that follows is the first one (entitledPurposes). Why? Because many years of experienceadvising institutional funds suggest that the chiefobstacle to investment success for many eleemosynary

funds is excessive diffusion of responsibility at thegoverning board level. To be sure, a shortage of timeor expertise at the governing board level can also besufficient conditions for failure, but having too manychefs in the kitchen is almost always a recipe fordisaster, especially if some of the chefs appear in thekitchen on a part-time basis only. In today’s wiredworld, trustees need not always be in the same room tohold meaningful discussions, and the ease with whichmeetings can be conducted via teleconference andvideoconference is making it increasingly easy forindividuals living in different cities (if not countries)to engage in joint decisionmaking. Interestingly, thesingle most successful investment committee of alltime meets telephonically far more often than it meetsface-to-face, a practice that admittedly is facilitated bythe committee’s very small size. (Mr. Buffett lives inNebraska, and his sole fellow committee member Mr.Munger lives in California.) Indeed, it may be thatimproving communications technologies arecompounding rather than reducing the problem ofhaving too many chefs in the kitchen: trustees whoseinterest in an institution’s investment process tends tobe episodic rather than constant may find it easier toinject themselves into the conversation if they can doso from the comfort of their own homes or offices.

* * * * *

MESSAGE IN A BOTTLE continued


4

Preface

Purposes. The purpose of this document is two-fold:(1) to codify the Foundation’s investment policies and(2) to codify the premises on which these policies rest.To make the policies more intelligible to first-timereaders, the premises (§I below) precede the policiesthemselves (§II). As noted below (see “Strengths”),the Foundation believes that investment committeesfunction best when they are kept small in size andwhen participation in committee deliberations isconfined to board members who are closely familiarwith the Foundation’s investment policies and practices.To the extent that members of the Foundation’s boardwho do not serve on its Investment Committee (“IC”)attend IC meetings, it is expected that they will haveread this document, recently and in its entirety, beforeplaying an active role in IC discussions.

Ongoing Review. To remain effective, theFoundation’s investment program must be responsiveto both the Foundation’s evolving needs and to changingexternal market conditions. Accordingly, it isimperative that the Foundation’s InvestmentCommittee review this document on an ongoing basisand make changes as needed. Material changes shallbe communicated to the Foundation’s Board of Trustees(“Board”).

Format. Within each section, boldface and italics areused to lend order to the paragraphs, as follows: theparagraph header for each topic appears in Boldface;the header for each subsidiary topic therein appears inItalicized Boldface; the header for each topic therein(“sub-sub-topic”) appears in Italics; and the headersfor any subsidiary paragraphs appear in RegularTypeface.

* * * * *

Part I. Premises

The major premises underlying the Foundation’sinvestment program can be usefully sub-divided intofour categories: premises respecting the Foundation’sidiosyncratic goals and competencies (§I.A); premisesrespecting the markets in which the Foundation’sassets are invested (§I.B); premises respecting methodsthat the Investment Committee should employ whendeploying the endowment across and within assetclasses (§I.C); and premises respecting the attributesto be sought in external managers (§I.D).

A. Foundation Goals and Competencies

Targeted Life. In contrast to many grantmakinginstitutions of comparable size, the Foundation doesnot seek perpetual life status. The Foundation ascribesmore importance to funding high-quality programsconsistent with its grantmaking priorities than it doesto the pursuit of perpetual life status for its own sake.The investment policies outlined below assume thatgrantmaking could cause the Foundation’s endowmentto become depleted fully by the late 21st century. Theactual depletion rate will depend on the magnitude andquality of grantmaking opportunities that arise andexternal market conditions.

Liquidity Needs. The Foundation is compelled bylaw to distribute an average of 5% of its assets perannum. By design, only a limited portion of theFoundation’s annual distributions are contractuallycommitted, and the Foundation stands prepared toreduce spending in constant dollars in order to preventthe forced sale of holdings at depressed prices. It alsostands prepared to distribute a substantial fraction ofits assets in a given year if compelling grantmakingopportunities arise. Accordingly, the Foundation hasa finite tolerance for illiquid investments (defined asinvestments not readily reducible to cash within 20business days).

Current Spending Policies and Commitments. Forcalendar year 1999, annual distributions plus expensesare expected to approximate 6% of gross assets at year-end 1998. The face value of unpaid future commitmentsto grantees approximates 10% of gross assets, or about8% of gross assets when discounted to present valueusing a nominal discount rate equal to the yield-to-maturity of 91-day US Treasury bills.

PI-Related Cash Flows. The Foundation is obliged tomeet capital calls from certain private investment(“PI”) managers that prefer to call down client fundson an as-needed basis only. In accordance with writtenagreements between such managers and the Foundation,the Foundation’s PI-related capital commitments aredefined over a multi-year time period. In practice,such commitments can sometimes be substantially ifnot fully met by redeploying funds generated by earlierPI commitments (cash distributions or sales of in-kinddistributions). However, the policies set forth in §IIassume that under worst-case conditions theFoundation’s external PI managers will suspend suchliquidations, obligating the Foundation to satisfy with


5


non-PI resources aggregate calls ranging as high as15% of total assets over a rolling five-year period.

Investment Time Horizon. Consistent with itstargeted life as defined above, the Foundation’stheoretical investment time horizon is long term. Itsactual investment time horizon is shorter than theorysuggests, however, for two reasons: first, theFoundation’s liquidity needs preclude theimplementation of a long-term buy-and-hold policyrespecting all of its assets; second, the Foundation’strustees and staff think it unwise to assume that theirinvestment decisionmaking will be entirely free ofbehavioral tendencies that cause most institutions’actual time horizons to be shorter than theory suggests(see Behavioral Pitfalls below).

Investment Objectives. The Foundation’s investmentobjectives must be consistent with the boundaryconditions set forth in this §I. These objectives arestated summarily below and discussed in greater detailin §II.

Primary Return Objective. The Foundation’s primaryreturn objective is to preserve the purchasing power ofthe unitized value of its endowment assets, net ofadministration and investment management costs, overrolling five-year periods. This goal is synonymouswith the pursuit of a time-weighted return onendowment assets that is at least equal to an inflationrate appropriate to the Foundation (the CPI or anagreed-upon alternative) measured over rolling five-year periods.

Secondary Return Objective. The Foundation’ssecondary return objective is to enhance the purchasingpower of the unitized value of its endowment assetsover rolling five-year periods. This goal is synonymouswith the pursuit of a time-weighted annualized returnon endowment assets that exceeds an inflation rateappropriate to the Foundation plus five percent net ofall costs, measured over rolling five-year periods.

Tertiary Return Objective. The Foundation’s tertiaryreturn objective is to avoid peak-to-trough declines inunitized endowment purchasing power exceeding 30%.

Underpinnings. The adoption of objectives expressedin unitized terms reflects an assumption that theendowment’s total market value could potentiallyshrink in accordance with the Foundation’s overallfinancial goal of potentially liquidating its assets. The

adoption of rolling five-year periods for assessment ofresults reflects a balancing of the financial goal with itsBoard’s perceived tolerance for unexpectedly poorresults. The five-year period for assessing results isintended as a floor but not a ceiling: its adoptionindicates that the Board ascribes little importance todeclines in unitized endowment purchasing power thathave persisted for less than five years, unless theyentail peak-to-trough declines exceeding 30%. The30% limit serves as a proxy for the Foundation’stolerance for changes in its financial condition that,however fleeting, would likely trigger fundamentalchanges in the scope and character of its grantmakingendeavors.

Other Attributes. The Foundation’s investmentpolicies must necessarily be sensitive to its distinctiveattributes, including the following:

Strengths. The Foundation derives a theoretical“edge” when investing (relative to most US-basedinstitutional investors) from at least four importantattributes:

Large Asset Base. At present, the Foundation’sendowment is large enough to permit it to surmountthe minimums imposed by external managers and toretain highly qualified outside consultants withoutincurring fees that are disproportionate in relation tothe Foundation’s total assets. This advantage coulderode over time if the absolute size of the endowmentdeclines due to a material acceleration in theFoundation’s grantmaking activities.

Small Committee Size. To avoid “lowest commondenominator” decisionmaking (i.e., decisions dictatedby the least expert person involved) as well as excessivediffusion of responsibility for portfolio results, theFoundation’s Investment Committee (“IC”) is small incomparison to peer group norms. At present, theCommittee comprises four voting members: theFoundation’s president, its chief investment officer,and two non-executive investment professionals. TheFoundation is committed to keeping the IC’smembership at five individuals or fewer under allcircumstances, including any staff members serving asvoting members of the Committee.

Low Concern for Reputational Risk. A fear ofunderperforming other investors (especially peerinstitutions) plays a central role in the stewardship ofmost eleemosynary organizations’ assets (especially



6

institutions engaged in ongoing fundraising) but is ofless relevance to the Foundation’s Board andInvestment Committee.1

PI Manager Relationships. The Foundation enjoys astrong relationship with several private equity firms inwhich its Investment Committee and staff haveconsiderable confidence. PI-related holdings andcommitments represent a material fraction of theFoundation’s total assets, reflecting the Committee’sconscious decision to bear the idiosyncratic risks thatsuch investments entail in exchange for the incrementalreturns that they are expected to generate relative tomore readily marketable equity investments. Aspreviously noted, the Foundation has entered into aseries of agreements with external PI managers thatoblige the Foundation to meet capital calls from themwithin defined limits. The Foundation can attempt toinfluence through discussions with such managers therate at which they liquidate investments made onbehalf of the Foundation and their other clients, but theFoundation has no direct control over the timing ofsuch liquidations.

Weaknesses. The Foundation’s investment programis potentially handicapped (relative to most US-basedinstitutions) by virtue of two attributes:

Private Foundation Status. The Foundation’s taxstatus renders it potentially vulnerable to at least twoforms of taxation to which many US-based tax-exemptinstitutions (e.g., ERISA funds and educationalendowments) are not subject: (1) unrelated businessincome taxes (“UBIT”) on debt-financed realtyinvestments and (2) excise taxes arising from a failureto fulfill Congressionally-mandated payoutrequirements. The Foundation is subject to other taxesalso (e.g., withholding on dividends derived fromcountries with which the US lacks fully exemptive taxtreaties). In practice, the Foundation’s tax statuscompels those stewarding its assets to assess carefullythe tax implications of all prospective investments. Italso compels those administering the Foundation’sinvestment program to design and implement tax-sensitive contractual arrangements with externalmanagers, including UBIT-related reportingrequirements that some prospective managers are likelyto find offputting.

Interplay between Tax Status and Large Asset Base.The tax-driven contractual and reporting requirementsjust described boost the “fixed costs” of certain typesof investments. These costs include one-time expensesincurred during the selection and negotiation phases ofeach investment, plus ongoing administrative andreporting costs. In light of these fixed costs, theFoundation must be careful to avoid makingcommitments that are too small to have a meaningfulimpact on its overall results. At present, the informalminimum commitment size is $30 million(approximately 3% of total assets), a requirement thatcan prove problematic when applied to managers withrelatively small asset bases: although the InvestmentCommittee has not adopted formal limits on thepercentage of a manager’s assets that the Foundation’scapital may represent, it recognizes the dangers inherentin commitments that are abnormally large in relation toa manager’s total assets and seeks to avoid commitmentsrepresenting more than 25% of such assets.

Behavioral Pitfalls. Those charged with stewardshipof the Foundation’s assets recognize that they are notimmune to the behavioral pitfalls that befall investors.A catalog of the pitfalls to which investors arevulnerable appears immediately below. The purpose


1 The Foundation is subject to prudent investor standards articulated bythe legislature of the state in which it is domiciled. While the “modern”prudent man rule represents a vast improvement over the “old” version,it still provides no meaningful guidance to well-intentioned trusteesseeking to choose among an almost infinite variety of plausible investmentoptions. The “modern rule” is flawed because it essentially states thatinstitutional investors are free to invest in anything that strikes their fancyas long as doing so will enhance their overall portfolios’ risk-adjustedreturns. When applying this test to a prospective investment, trusteestypically do that which a judge or jury will assumedly do if the investmentgoes sour and it is subsequently challenged in court: they analyze how theaddition of the asset in question would have affected the utility of the pre-existing policy portfolio on a historical basis. With this decision rule, ofcourse, the better an asset has performed historically, the more merit thereis in adding it to an institution’s asset mix today. In other words, the“modern” prudent man rule creates a not-so-subtle incentive to “buyhigh,” as in small cap stocks circa 1983, real estate circa 1986, Japanesestocks circa 1989, emerging market stocks circa 1993, and so on.Ironically, all of the asset classes just mentioned are now objectively“cheap” in relation to historical valuation standards. They have become“cheap” for fundamental reasons, of course, but also because backward-looking assessments of asset class characteristics have caused differenttypes of “alternative investments” to supplant these first-generation“alternatives” as the logical repositories of “smart money,” the mostconspicuous examples being “private equity” broadly defined and so-called “absolute return” strategies. One expert on fiduciary standards hasstated that a careful review of every published legal opinion in his fieldsince 1800 reveals not a single instance in which a trustee of aneleemosynary organization has been found liable for flawed investmentdecisions — excepting cases of self-dealing. In other words, the countlesstrees and the many hours of legislators’ and attorneys’ time expended inpromulgating the “prudent man” rule in its various guises could have beensaved with a simple three-word injunction: “no self-dealing.”

7


of this catalog is to remind members of the Foundation’sBoard, Investment Committee, and staff of the ever-present danger of such traps:

Overreaching. The Foundation’s investment programis geared consciously toward the avoidance of a pitfallto which institutions with large asset bases are especiallyvulnerable: attempting to play many games adequatelyrather than a few games well. “Games” is not used herein a pejorative sense but rather as shorthand for themyriad tasks that institutional investing necessarilycomprises. In an ideal world, the Foundation couldadd human resources as needed (at the trustee or stafflevel or both) to exploit any and all return-optimizationopportunities that arise. But there are concrete limitsto how large the Investment Committee can expandwithout producing both lowest common denominatorthinking and an excessive diffusion of trusteeresponsibility. Moreover, there are real limits to howlarge the staff can expand without displayingbureaucratic tendencies, and it is unreasonable tosuppose that the Foundation can attract and retain first-rate investment professionals if these individuals aresaddled with excessive administrative burdens.Consequently, the Foundation must routinely declineinvestment opportunities that comport with itsinvestment objectives when viewed in isolation,mindful that the returns expected from any newinvestment must be adjusted downward to reflectexogenous costs, i.e., costs arising from the allocationof scarce resources to one “game” as against all other“games” that might conceivably be “played.” Indetermining which “games” to play, the Foundationmust assess both the magnitude of the opportunitybeing evaluated and the probability that the opportunitycan be exploited profitably. Differently put, it musttake care to allocate scarce human resources (at boththe staff and trustee level) based on probability-weighted assessments of likely outcomes, avoiding“games” that are relatively difficult to win unless theirpotential payoffs are sufficiently large.

Implications for Asset Allocation. As noted in §I.Cbelow, the Foundation’s preferred approach to assetallocation assumes that neither the InvestmentCommittee nor staff can accurately and consistentlyforecast the near-term direction of capital markets.The approach also assumes that macroeconomicconditions that either have already unfolded or arereasonably certain to unfold over the near term arealready reflected in securities prices. Accordingly, the


Investment Committee and staff devote only as muchtime to discussing such conditions as is needed toexercise intelligently the discretion they possess toengage in asset substitution, as that term is defined in§II.C below.

Implications for Active Manager Selection. To theextent that institutional funds engage in “overreaching,”they tend to do so most frequently with respect toactive manager selection. Selecting superior activemanagers is in important respects more difficult thanselecting winning stocks, for two reasons: (1) thepeople making the decisions (i.e., InvestmentCommittee members or staff working under theirsupervision) have far less information about managersthan investors have about publicly traded companiesand (2) these people must balance the conflictingpropaganda of managers (many of whom are quiteglib) against the only facts available to them, i.e.,managers’ past performance. Accordingly, fiduciariesor their agents invariably base manager selections onavailable facts rather than unproven and unprovablemarketing claims, causing them to favor recent winnersover recent laggards. Managers’ marketing claims aregenerally unprovable because it takes a long time toprove (statistically) that favorable results reflect skillrather than luck (see exhibit below). Given thesehighly unfavorable boundary conditions, most managerchanges are a mistake. Firings are generally a mistakebecause clients usually bail out just before a manager’s

54

18

4 2 0.170

20

40

60

Stocks vBonds

EAFE ex Japanv Japan

5 CountryDecisions

10 SectorDecisions

100 StockDecisions

Num

ber o

f Yea

rs 75% Confidence Level

120

408 4 0.42

0

40

80

120

160

Stocks vBonds

EAFE exJapan v Japan

5 CountryDecisions

10 SectorDecisions

100 StockDecisions

Num

ber o

f Yea

rs 95% Confidence Level

Years Necessary to Prove Ability to Add Value

Assumptions: (1) Returns are normally distributed.(2) Decisions are correct 60% of the time.

Concept Credit: Grantham Mayo Van Otterloo & Co.


8

style turns, and hirings are generally a mistake becauseclients are chasing recent successes. The underpinningsof such behavior are identified in the next twoparagraphs.

Herding. This is the natural human instinct to movewith the crowd. In an institutional setting, it manifestsitself in an aversion to “reputational risk,” which inlayman’s terms means simply the risk of being wrongand alone. Given most investors’ extreme discomfortwith being wrong and alone, investments entailing ahigh degree of reputational risk tend to providedisproportionately high expected returns.2

Representativeness. This pitfall manifests itself wheninvestors assign too much weight to recent events. Forexample, polls taken over the last several years showthat an alarmingly high percentage of individualinvestors think that the broad US stock market willgenerate 15% or higher returns over the long term. Thebroad US stock market has indeed produced such loftyreturns in recent years, but in order to continue doingso from today’s lofty price levels, either profitabilitylevels or valuation measures (e.g., price/earnings ratios)would have to soar to levels that strain credulity.Representativeness manifests itself repeatedly withrespect to so-called alternative investments, creatingbig problems for institutional investors who mistakenlybelieve that strategies which have produced very strongreturns in recent years can continue doing so even ifthey become infused with shockingly large amounts offresh capital. Recent examples would include thereturns expected from emerging market stocks acquiredat the zenith of their popularity in 1993 or fromdistressed securities entering 1998.

Risk Aversion. Risk aversion entails a distaste foruncertainty per se, including unexpectedly favorableoutcomes. Thus, when confronted with a choicebetween a certain gain (e.g., a 5% return) and anuncertain outcome whose expected value exceeds thecertain gain (a 50% chance of a 30% gain coupled witha 50% chance of a zero gain), most investors opt for the


certain gain. Risk averse decisionmaking is not alwaysin the best interest of an institutional fund with a verylong-term investment time horizon.

Loss Aversion. Although much of the academicliterature on investing assumes that investors are riskaverse, the typical investor is more properly describedas loss averse. Loss aversion entails a distaste forcertain failure. Thus, when confronted with a choicebetween a certain loss (e.g., closing out a positionentailing a 10% loss) and an uncertain outcome whoseexpected value is even more detrimental (a 50% chanceof a 15% loss coupled with a 50% chance of a zeroloss), most investors opt for the uncertain outcome.Loss averse decisionmaking and a related phenomenonknown as “pride of ownership” tend to inhibitdecisionmakers from redeploying capital fromsuboptimal investments to potentially more productiveones.

Anchoring With Reference Points. The most naturalreference point for most investors is their cost basis.As noted previously (see Loss Aversion), this referencepoint can produce highly suboptimal decisions. So toocan the anchoring that occurs with respect to prospectiveinvestments, i.e., assets not already held in the portfolio.In the latter context, investor views tend to be anchoredin historical data. For example, because US stocks hadtraditionally performed poorly for multi-year periodsfollowing their initial display of dividend yields below3%, many solidly anchored investors reduced theirexposure to US stocks when the dividend yield on theS&P 500 dipped below 3% in the mid-1980s (andagain in the mid-1990s). This particular anchor hasproven extremely costly.

Framing. This is the tendency to try to make acomplex world simple by attaching overly broad andconvenient labels to data or ideas that merit moreheterogeneous treatment. A topical example: treatingthe strategies and tactics employed by commingledinvestment vehicles operated pursuant to specifiedcontractual arrangements (i.e., hedge funds) as a distinct“asset class.”

B. Markets

No Free Lunch. Given the relatively free flow ofcapital within and among nations in the modern globaleconomy, the Foundation’s operative assumption is

2 A vivid recent example would be Korean stocks at the end of 1997 —assets that had performed so poorly for so long that most fiduciaries (andtheir agents) would not think of holding them, let alone putting freshmoney into them. As the Korean economy moved into a more stable statein 1998, investors realized that Korean stocks were excessively depressedin price, and the injection of modest amounts of fresh capital into themarket made Korea the best-performing market in the MSCI database in1998.

9


that all investments produce roughly equivalent risk-adjusted returns. Differently put, the Foundationattributes differences in the returns produced bydifferent types of investments to differences in thetypes and degrees of risk that they entail. Consistentwith this mindset, the Foundation devotes substantialresources to determining the types and degrees of riskthat each investment entails and to determining whetherthe Foundation is well-suited to serve as the marginalbuyer (or seller, in the case of prospective short sales)of the asset in question.

Market “Efficiency.” The statement that “allinvestments produce roughly equivalent risk-adjustedreturns” is not intended as an endorsement of the viewthat capital markets are perfectly or even approximately“efficient” as that term has traditionally been defined.Indeed, to the extent that “efficiency” has traditionallybeen defined as returns that are proportionate to eachasset’s price volatility (typically measured overrelatively brief holding periods such as 12 months), theassumption about markets advanced in the precedingparagraph is actually intended as a refutation of theprevailing view. As noted below, short-term pricevolatility is one valid measure of investment hazard,but it is by no means the only form of risk that theFoundation expects to be rewarded for bearing. Inshort, the Foundation’s investment program assumesthat diligent research will enable the Foundation toidentify investments that generate superior returnsafter being adjusted for short-term price volatility.

Almost-Free Lunch. The most plausible exception tothe “no free lunch” rule is the Foundation’s potentialexploitation of opportunities arising from its status asa very high grade borrower whose investment returnsare essentially tax-exempt. However, as the Foundationand other tax-exempt investors have discovered,exploiting such opportunities through the conscioususe of leverage entails costs that go beyond interestexpenses, including the inherently unquantifiableopportunity costs associated with investment choicesthat would have been better researched or betterexecuted had those responsible for such choices beenless preoccupied with debt service concerns.Accordingly, the investment guidelines set forth belowpermit the Investment Committee to employ leveragewithin specified bounds. The guidelines also encouragebut do not require the Committee to include financingcosts in the determination of whether a specificsubstitute asset (i.e., any asset or strategy other than


one reflected in the Policy Portfolio) merits inclusionin the actual portfolio, and also to earmark borrowingsto the extent needed to facilitate the Committee’sselection of an appropriate gearing ratio.

Rigorous Standard. Earmarking borrowings entails aheavier burden of proof against debt-financedinvestments than non-earmarking because expectedreturns receive no boost from “the portfolio effect,”i.e., the tendency of combinations of non-correlatedassets to produce long-term returns that exceed theweighted average of their individual returns. Asdiscussed below, to the extent that portfolios compriseilliquid assets whose weights cannot be rebalancedreadily, or assets whose correlations tend to soar underworst-case conditions, “the portfolio effect” has beenvastly oversold. It has been oversold because theincremental returns that “the portfolio effect” producesin computer simulations evaporate if real-worldconstraints (be they legal or behavioral) precludedisciplined rebalancing.

Risks. There are just a few types of risk that are trulygermane to the Foundation and that it demands to berewarded when bearing.

Inflation Risk. The most worrisome form of investmenthazard that the Foundation confronts is inflation risk,i.e., the risk that an asset or portfolio of assets will nothold their value in real or inflation-adjusted terms. Asused here, “long term” means the multi-decade timehorizon defined above (see Investment Time Horizonin §I.A).

Caveats. For perpetual life institutions, equities broadlydefined are the least risky asset because they are themost reliable way of maintaining endowmentpurchasing power. For investors with potentiallyfinite time horizons, assets whose prices fluctuate(including equities) can be very risky because suchinvestors may need (or elect) to sell them at depressedprices. While mindful that equities broadly definedhave outperformed bonds over the very long term, theFoundation is also mindful that equities broadly definedhave been known to produce negative real returns forextended time periods. Importantly, the inflationaryconditions to which such periods of disastrous returnscan be attributed (e.g., the –17.7% inflation-adjustedcumulative return produced by the S&P 500 during the10 years ending in 1982) also tend to wreak havoc onthe right-hand side of the Foundation’s balance sheet:


10

to maintain their effectiveness during an extendedinflation, grant payments must be ratcheted upward innominal terms, as must staff salaries and otheradministrative expenses. In short, very high rates ofunanticipated inflation place endowed institutions in aposition not unlike an object tossed into a compactor:falling stock prices place ferocious pressure on theleft-hand side of their balance sheets (i.e., assets) whilesoaring costs create countervailing pressures in theopposite direction. To hedge against the forced sale ofequities at depressed prices under highly inflationaryconditions, a defined portion of the Foundation’sportfolio is invested in inflation-indexed bonds (oralternate inflation hedges) at all times.

Deflation Risk. A second and very different form ofcatastrophic risk that can befall the Foundation isdeflation. Unfortunately, the assets that most reliablyhedge long-term inflation risk (i.e., equities broadlydefined) are distinctly unhelpful under deflationaryconditions. Indeed, equities tend to performdisastrously under such conditions. That said, deflationarguably is less worrisome than hyperinflation toendowed institutions because liabilities tend to fall innominal terms when deflation strikes, mitigating thedamage sustained on the asset side of such institutions’balance sheets. However, deflation can cause equityvalues to plummet more rapidly than the cost of goodsand services that the Foundation seeks to purchase(directly or via grants). To hedge against the forcedsale of equities at depressed prices under deflationaryconditions, a defined portion of the Foundation’sportfolio is invested in high quality, non-callable,long-term bonds at all times.

Short-Term Price Volatility. This is germane for tworeasons. First, the Foundation has liquidity needs thatare potentially difficult to satisfy through currentincome. The more short-term volatility an asset orasset class displays, the more vulnerable the Foundationis to forced sales at depressed prices to meet cash flowrequirements. Second, quite apart from liquidity needs,short-term price volatility can produce “whipsaw” —the reflexive sale of holdings at depressed prices not tomeet cash flow needs but to satisfy the risk tolerancesof the persons charged with stewarding theFoundation’s endowment. Whipsaw is especiallyworrisome when fiduciary risk tolerances arethemselves unstable, a phenomenon that cannot beruled out even in the absence of trustee turnover.


Illiquidity. Although liquidity risk is largely subsumedby price risk — i.e., “there is a market-clearing pricefor everything” — the Foundation deems it useful todistinguish between publicly traded assets whoserealizable value is subject to short-term swings inmarket sentiment and privately traded assets whoseimmediately realizable value is constrained by therelatively high transaction costs that are necessarilyincurred when their ownership changes hands.3 Thissensitivity to potentially penal transaction costs isreflected in the policy guidelines set forth in §II.

Informational Risk. The Foundation does not expectto be rewarded for bearing uncertainty attributablesolely to its own failure to collect and assess intelligentlylegally obtainable information. But it does expect tobe rewarded for bearing uncertainty that cannot beeliminated through disciplined research and analysis.The corollary is crucially important: given the intensecompetition among investors for abnormally profitableinvestments, the Foundation assumes that informationwhich is readily available to investors is alreadyreflected in asset prices. This is especially true withrespect to (a) information respecting forecastable eventsthat have not yet occurred and (b) historical returns onasset classes. Putting both points differently: (a) it isthe true surprises that move markets on the margin and(b) by the time investors accumulate sufficient evidencesuggesting that an asset class (or investment approach)produces superior returns, any abnormal profitsassociated with such opportunities have already beenmade.

Hedging Costs. Disaster hedges such as thosecontemplated by the policies set forth in §II aresometimes so mispriced that the “insurance” theyprovide has a “negative cost,” i.e., investors get paid tobecome insured. An example: in 1981, long-term

3 There is an important nexus between liquidity risk and “whipsaw risk”(as that term is defined in the preceding paragraph of the main text). Bytheir very nature investments that entail a high degree of liquidity riskentail a low degree of whipsaw risk. In other words, if an institutioncannot bail out when things go unexpectedly poorly over a given periodof time, it is much less vulnerable to getting whipsawed. That does notmean the Foundation should invest exclusively in illiquid investments.And it does not mean that an enterprising external manager can make anotherwise liquid investment more attractive to the Foundation by subjectingit to a lock-up. It simply highlights the fact that the Foundation’s chieftask is to make informed judgments about appropriate tradeoffs betweenrisk and return. If extreme liquidity constitutes a drawback under certainnarrow conditions, then illiquidity constitutes a virtue under the sameconditions.

11



Treasury bonds (the classic deflation hedge) yieldedmore than 14%, an unprecedentedly high nominalyield that provided the potential for rich capital gainsif the Federal Reserve succeeded in bringing inflationunder control (as it subsequently did) and significantprotection (via reinvestment of coupon income) if itdid not. There is an important corollary to the principlethat investors can occasionally get paid to becomeinsured: such hedges can also become so overpricedthat investors who acquire them at peak prices canwind up worse off than investors who lack such“insurance” — even if the event being insured againstoccurs. An example of the latter phenomenon: investorswho acquired energy-related properties for inflation-hedging purposes at the peak of energy prices in 1980were going to do very poorly even if inflation hadcontinued spiraling upward. Of course, it did not.

Equity Real Estate as an Inflation Hedge. Theabsence of a permanent allocation to real estate in theaccompanying guidelines reflects two assumptionsabout this asset class: (1) that the Foundation’s taxstatus reduces materially expected after-tax returnsfrom debt-financed equity real estate and (2) thatcertain forms of equity real estate may not providemeaningful diversification versus other equity-orientedassets during inflation-induced bear markets in stocks.Accordingly, the Investment Committee’s operativeassumptions respecting equity real estate are (1) thatany new commitments to equity real estate will bepreceded by a careful assessment of after-tax returnsand (2) that they will be subject to the stringentapplication of the Asset Substitution criteria set forthin §II.C.

Commodities as an Inflation Hedge. Theaccompanying guidelines exclude a permanentcommitment to commodities for reasons similar tothose underlying the exclusion of a permanentallocation to equity real estate. Accordingly, thepurchase (or retention) of commodity-related holdingsalso shall be subject to the stringent application of theAsset Substitution criteria set forth in §II.C, withspecial attention being paid to supply and demandforces within commodity markets that could causesuch holdings to under- or outperform the marketbenchmark for the Foundation’s inflation hedgingsubsegment. As a general rule, the Foundation willeschew investments in raw commodities but standswilling to acquire commodity-related assets to whichvalue-enhancing strategies and tactics can be applied

(e.g., energy reserves held by deftly managedoperators).

C. Asset Allocation Approaches

Overview. This subsection has a twofold purpose: (1)to describe summarily the approach that the InvestmentCommittee employs when formulating asset allocationguidelines and (2) to codify the reasons why it does notrely more heavily on alternate approaches favored bymany large institutional funds. To make theCommittee’s favored approach more intelligible, thecritique of alternate methods appears first.

Mimicry Decried. The ultimate measure of theappropriateness of a given asset mix for a givenfoundation is whether it will produce the maximumreturn for whatever level of risk the Foundation’strustees are willing to bear. But few if any trusteegroups can gauge accurately their tolerance for poorresults until those results actually roll in. The idea ofan “ideal” asset mix for foundations with differentgoverning boards is thus an impossibility. Whenadded to the Foundation’s relative indifference to peergroup comparisons (see §I.A above), this fact obviatesthe need for the Foundation to mimic intentionallypeer institutions’ policies and practices.

“Efficient” Portfolios Decried. The most popularmethod used by institutional investors whenformulating asset allocation guidelines is mean-variance analysis (“MVA”), the aim of which is toidentify “efficient” asset mixes. The aim ischoiceworthy — the Foundation does indeed seek toearn the highest possible returns for any combinationof risks that it elects to bear— but the means typicallyemployed in pursuit of this aim are defective in multiplerespects. These defects are catalogued here to remindreaders why the Foundation does not rely more heavilyon MVA or related approaches when making assetallocation decisions.

Partial Critique Only. What follows is by no means acomprehensive catalog of MVA’s perceived defects.Rather, it is a partial critique that discusses only thosedefects of MVA that are displayed by the computer-based model that the Foundation’s external investmentconsultant has made available for the Foundation’suse. In contrast to conventional mean-varianceoptimizers, this model has the important advantage of


12

focusing not on a single undifferentiated holding periodbut rather on multiple time periods and of facilitatingthe assessment of alternate portfolios’ riskiness usingclient-specified definitions of risk. For modelingpurposes, the Foundation has traditionally defined riskin two ways: first and foremost, as potential shortfallsin terminal wealth below a given threshold; andsecondarily, as differences between the Foundation’sreturns and the returns realized by institutions pursuingwhat are assumed to be more conventional policies. Inits most recent application, the model was told toweight the first definition 10 times more heavily thanthe second, in keeping with the Foundation’s relativeindifference to peer group comparisons as describedabove. The model in question is quite elegant, especiallyin its capacity to analyze risk through prisms morerobust than mean-variance, but an elegant dwellingbuilt on an inherently shaky foundation provides nomore shelter when earthquakes strike than a cruderabode built on the same shaky foundation. Thefoundation on which the model rests is defective in thefollowing respects:

Clairvoyants Not Available. The chief defect of themodel is that it presupposes more forecasting acumenthan the Foundation or any advisors available to it arelikely to possess. To elicit output from the model, itsuser must specify the expected returns, standarddeviations (“variances”), and correlations(“covariances”) of the asset classes being analyzed.Importantly, empirical studies by leading statisticiansworking in the investment arena indicate that thereturn forecasts are roughly 10 times as important (interms of indicated asset mixes) than forecastedvariances, with the latter in turn being roughly twotimes as important as forecasted correlations.Differently put, institutions that turn to MVA modelsin an effort to sublimate if not eliminate forecasting inthe policy formulation process are doing nothing of thesort.

Speciousness of Historical Return Data. It is unwiseto rely solely on unadjusted historical returns as inputs,because such data ignore the dynamism of capitalmarkets. Simply stated, the current price of an asset isalways more important than historical averages.Moreover, studies extolling the virtues of particularinvestments tend to appear at or close to secular peaksin the returns on such investments. Examples abound:small stocks in the US (horribly expensive by 1983,when the first big wave of studies was published


extolling their long-term virtues); real estate in themid-1980s (same story as small stocks); developedmarket foreign stocks in the late 1980s (ditto); andemerging market stocks circa 1992-93 (samephenomenon).

Dangerous Vacuum. The return expectations onwhich investment policies formulated usingoptimization models rest tend to be formed in a vacuumwithout regard to the fact that countless other institutionsare engaged in precisely the same exercise of forecastingexpected asset class returns. (Logisticians refer to thisas the “fallacy of composition.”) Consensusexpectations are not always wrong, but they are seldomright with respect to asset allocation opportunitieswhose aggregate size is a fraction of the institutionalcapital pursuing such opportunities in the mistakenbelief that they can absorb vast sums of fresh capital(e.g., US venture capital circa 1983-84, emergingmarket stocks circa 1992-93, fixed income arbitragecirca 1997-98). When using the optimization modelsupplied by its external investment consultant, theFoundation makes the assumption that all types ofmarketable stocks held by the Foundation will producethe same return. This mitigates the problem of assumingthat relatively small pockets of opportunity (e.g.,emerging market stocks) can accommodate large sumsof fresh capital without undermining their assumedreturn advantage, but it introduces a different form ofarbitrariness into the modeling process that is alsounsettling.

Dismal Track Record. The forecasting errors alludedto above cannot be attributed solely to the paucity ofdata respecting non-traditional assets with whichforecasters were confronted when these novelingredients were initially tossed into the optimizationstew. After all, the forecasted returns of the more“traditional” assets analyzed in such studies also provedwildly inaccurate, albeit mostly on the low side. Indeed,foreign stocks (developed as well as emerging) are theonly major asset classes that have produced returns inthe 1990s below those forecasted by MVA users whenthe decade began. (Commodities represent a thirdexception.) But the forecasts of other asset classes’behavior have proven equally inaccurate, a factobscured by the generally benign character of theforecasting errors in question. Trustees find it hard toget upset when the major markets in which they’veinvested perform much better than expected (e.g., USstocks and bonds over the last 16 years). But they are

13



far less tolerant of materially smaller forecastingerrors when results are much worse than expected,especially when the assets purchased for returnenhancement reasons produce negative returns (e.g.,emerging market stocks 1993-present).

Speciousness of Other Inputs. The illogic of relyingsolely on historical return data is rivaled by the illogicof relying on historical volatilities and correlations.The latter variables are no less dynamic than returnsand hence no less difficult to forecast, even over multi-year time horizons. Again, examples abound, the mosttopical being the wild inaccuracy of the expectedbehavior of emerging market stocks when these wereadded to institutional asset mixes in the early- to mid-1990s. Commodities represent a second example, notprimarily because commodity returns have beendreadful of late but because commodities performedso poorly during the two biggest downdrafts in worldstock markets since commodities became aninstitutionally accepted holding about a decade ago(i.e., the 1990-91 bear market and the briefer 1998stock market swoon). More specifically, the correlationbetween commodities and marketable stocks wasassumed to be negative, but commodities fell in priceduring both of the market downdrafts just mentioned,and dramatically so during the summer of 1998. Ofcourse, if inflation fears rather than recession fearstrigger the next big downdraft in global stock prices,commodities might indeed display a negativecorrelation with stocks. As for equity real estate —another asset class traditionally viewed as an “inflationhedge” — this too performed horribly during the bearmarket in stocks in 1990-91, providing virtually nodiversification vis-à-vis publicly traded equities whensuch diversification was needed the most. As Chow,Jacquier, and Kritzman have demonstratedconvincingly,4 asset class correlations have a nastybut proven tendency to soar when financial marketsare confronted with a major “shock” (e.g., Iraq’sinvasion of Kuwait in 1990, Russia’s default in August1998, etc.). If the chief purpose of optimizationmodels is to identify portfolios that will generatesatisfactory returns within acceptable volatilityconstraints, then institutions relying on such modelsshould use correlations that reflect worst-case

conditions, as distinct from “normal” conditions. Alas,highly diverse portfolios that appear “efficient” (i.e.,embody an acceptable balance between return andrisk) under “normal” conditions invariably embodyintolerable levels of downside risk when analyzedunder worst case conditions (deflation orhyperinflation).5

Inherent Contradiction. As a means of selectingrational asset allocation guidelines in an uncertainworld, MVA and related optimization methods do notwithstand logical scrutiny. Excepting whatever valueinvestors might derive from combining assets that areunarguably non-correlated under worst case conditions(e.g., long-term Treasuries vs. unseasoned stocks underdeflationary conditions), investors seeking meaningfulguidance from their MV-based models cannot escapeone ugly fact: if they truly believe their own returnforecasts, they should load up on the asset(s) with thehighest assumed returns, subject only to liquidityconstraints. (Investors who are truly clairvoyantarguably can skip the back-end liquidity check, mindfulthat lending institutions will furnish as much liquidityas they need to finance their demonstrably winningways.) Of course, most investors lack such confidencein their own return forecasts, else their portfolioswould be more concentrated, and they would pay littleif no heed to MVA. Lacking complete confidence intheir return forecasts, the investors we’re discussingtake a step that is as natural (in light of their insecurities)as it is perverse: they hedge their return forecasts byconstructing diversified portfolios that seem “efficient”based on a different set of equally suspect assumptions.These assumptions include the forecasted volatilitiesand correlations of the asset classes deemed eligiblefor purchase, their own risk tolerances (“utilityfunctions”), and — critically — their own ability andwillingness to rebalance portfolios in as disciplined amanner as their computer-based models presuppose.

Unrealistic Rebalancing Assumptions. A key premiseunderlying MVA and related optimization methods

4 “Optimal Portfolios in Good Times and Bad,” Financial Economist, No.272-1, December 1998. This paper was summarized nicely in theDecember 15, 1998, edition of “Economics and Portfolio Strategy,”published by Peter L. Bernstein, Inc.

5 Even the most ardent proponents of optimization models concede thatsuch models work imperfectly at best when equity real estate is includedas an eligible asset. This is true under even so-called normal marketconditions because real estate is traded on a discontinuous basis, renderingprice series insufficiently robust. Indeed, some leading consultants refuseto include real estate in optimization studies, preferring instead to relegateit to “alternative investment” status. In theory, this means that real estatehas no permanent place in the indicated policy portfolio even if it doeshave a more or less permanent place in the actual portfolio.


14

(including penalized shortfall models such as the onecurrently employed by the Foundation) is that assetmixes can and will be rebalanced periodically. Indeed,much of the “efficiency” inherent in portfolios that lieon the “efficient frontier” derives from the inherentlycontrarian recycling of funds from asset classes thathave outperformed to those that have lagged. Butseveral of the asset classes that make “efficient”portfolios choiceworthy on paper are inherently illiquid,private equity being the prime example. To surmountthis obstacle, many computer-based asset allocationmodels permit users to specify rebalancing costs foreach asset class being analyzed. This mitigates theproblem but hardly eliminates it: an institution thatseeks to reduce private equity exposure through meansother than the normal maturation of such holdings(e.g., through sales of partnership interests on thesecondary market) will quickly discover that its accessto attractive private equity opportunities has beenmaterially reduced due to general partners’ concernsabout its staying power. Quite apart from whether aninstitution is able to rebalance its assets in as disciplineda manner as most MV-based models presuppose,experience suggests that all but the most intrepidinstitutions are unwilling to do so. They are unwillingbecause rigorous rebalancing sometimes requiresselling assets that have been moving sharply higherwhile simultaneously buying assets that have beenplummeting in price (e.g., topping off emerging marketstocks in late 1998 via sales of US equities). TheFoundation could potentially constitute an exceptionto the latter rule, but before adopting policies thatpresuppose such intrepidity the Foundation shouldmake triply sure that all parties who enjoy a de facto orde jure veto over investment decisions have internalizedcontrarian principles.

Potentially Costly Distraction. Mindful of the aboveproblems with computer-based asset allocation models,many institutions and their consultants nonethelessdevote much time to their use. They do so in the beliefthat scrutiny of multiple model runs, each reflectingdifferent inputs and perhaps also time horizons, canshed important light on the tradeoffs inherent inchoosing one asset mix over competing alternatives.Perhaps it can, although in practice most institutionsspend far more time tweaking forecasted returns thanthey do tweaking forecasted volatilities or correlations— a potentially dangerous practice given theimportance that most asset allocation models ascribeto asset pairings displaying low or negative correlations.


But the greater danger inherent in such scenario testingis that it becomes preoccupying, consuming researchand discussion time that is disproportionately large intwo respects: first, in relation to the low accuracy of theassumptions underlying computer-based optimizerswhen viewed in hindsight; second, in relation to themuch greater potential value-added from applying thesame scarce resources to manager selection and otherpolicy or strategic issues that the fiduciaries involved(especially Investment Committee members) routinelyconfront.

False Axiom. The typical investment committeedevotes most of its time to asset allocation in theaxiomatic belief that this aspect of the investmentplanning process has the greatest influence on a fund’slong-term performance relative to its peers. It can,potentially, but only if the asset mix ultimately adoptedis materially different from the average mix employedby peer institutions. Alas, to wind up with a highlyunconventional policy portfolio, a trustee group that isheaded down this path typically spends an inordinateamount of time agonizing over model inputs andconstraints, which necessarily differ from those usedby other investors. Having stuck their necks out in thismanner, and perhaps exhausted most of the timeavailable for research and discussion, our hypotheticalgroup of maverick trustees often finds it difficult todeliver the goods, i.e., to deploy the endowment in amanner that will render their return forecasts reasonablein hindsight. This is especially true with respect to so-called alternative assets (a slippery term that used heremeans anything other than publicly traded US stocksand bonds). For better or worse, one cannot buy manyalternative assets on an indexed basis, private equityand absolute return-oriented investments providingtwo concrete examples. To actually realize the loftyreturn goals for these two asset classes that are reflectedin most “state of the art” asset allocation studies,trustee groups (or hired guns working under theirsupervision) must devote substantial amounts of timeto manager selection and evaluation within these twoasset classes. This is easier said than done, because theprocess of adopting an unconventional policy portfoliooften leaves scant time available for the latter tasks.

Lamentable Fact. The lamentable fact is that computer-based approaches to asset allocation do not eliminatethe need for fiduciaries (or their advisors) to makesubjective judgments about the future returns,volatilities, and correlations of the asset classes and

15



subclasses being considered for purchase. Indeed,they actually amplify this need, and they put fiduciariesin the untenable position of attempting to forecast theinherently unforecastable. No sensible trustee wouldclaim an ability to forecast asset class returns over aone-day or even one-year holding period becauserealized returns over such short-term periods can beaffected materially by numerous imponderablevariables. Conversely, no sensible trustee would relyheavily on MVA or related methods when formulatinginvestment policy for an endowment whose investmenttime horizon is indeed perpetual and whose spendingrate lies wholly within its board’s discretion: suchreliance would be misplaced under these conditionsbecause model inputs geared toward the medium-termtime horizons that actually dominate most trustees’thinking (i.e., 2–3 years in most cases, 5-7 years iftrustees are abnormally patient) become nonsensicalwhen the scenarios being analyzed are truly long term(e.g., 30-50 years or more). They become nonsensicalbecause correlations among what are today regardedas distinct “asset classes” (e.g., foreign vs. US stocksor US stocks vs. US real estate) rise as the measurementperiod lengthens, and returns migrate in a manner thatcasts doubt on modern analysts’ well-intentioned effortsto parse financial markets into a half-dozen or more“asset classes.” In short, the financial landscape isboth simpler and more complex than most computer-based approaches to asset allocation imply: simplerbecause what appear to be a myriad of distinct assetclasses can be prudently reduced to just three segmentsfor long-term planning purposes (as illustrated in §II)and more complex because the returns that anyinstitution can legitimately hope to realize from eachof these segments depend heavily on the specificstrategies and managers employed within them.

Distributions vs. Point Estimates. Although theconcerns raised in the preceding paragraph areespecially apt with respect to asset allocationapproaches based on point estimates of expectedreturns, they are highly germane to the more commonpractice of forecasting return distributions (i.e.,attaching probabilities to a vast if not infinite spectrumof potential outcomes). In academic parlance, pointestimates are referred to as “means” — shorthand formean returns. These means are accompanied byforecasted variances or standard deviations to produceexpected return distributions. Unfortunately, the actualdistribution of returns for most asset classes and

subclasses tends to be disturbingly unstable — “non-stationary” in academic parlance.

Proposed Use of Computer-Based OptimizationModels. The Foundation will continue to use computer-based optimization techniques (via its externalconsultant’s model), but for descriptive rather thanprescriptive purposes. In other words, it will use MVAto analyze the potential implications of asset allocationand manager selection choices arrived at by moreintuitively appealing means. Put yet another way, withthe exception of passive investments that will trackclosely the segment and subsegment benchmarksidentified in §II, no strategy or manager will beemployed unless it has the potential to satisfy thecriteria for Substitute Assets set forth in §II — regardlessof the asset or manager’s perceived utility when viewedthrough the prism of MVA and related optimizationtechniques. The Foundation consciously relegatessuch techniques to this subordinate rather than supremerole in the investment planning process because itsInvestment Committee and staff do not have sufficientconfidence in the assumptions on which computer-based optimization rests. To keep the resources devotedto such optimization reasonable in relation to its limitedutility, the Foundation will confine its use to thescrutiny of probable downside volatility under worst-case conditions, defined as economic scenariosentailing either deflation or very high rates ofunanticipated inflation. The chief if not sole purposeof such “stress testing” will be to ensure compliancewith the tertiary return objective specified herein(avoiding 30% or greater peak-to-trough declines inthe endowment’s unitized value).

Focal Point. Consistent with the assumption that time(at the staff and Investment Committee level) is likelyto be the most important determinant of theFoundation’s realized returns within each segment,the policies outlined below assume that the Foundationwill seek index-like returns from its two hedgingsegments (deflation-hedging bonds and inflation-hedging hard assets) in order to devote the maximumavailable manpower to determining which assets shouldbe held within its Total Return segment (equitiesbroadly defined). The Committee stands prepared tomake judicious use of so-called Substitute Assetswithin the Hedging subsegments as well as the TotalReturn segment, but its chief task (aided by staff) is toidentify and exploit opportunities to deploy the TotalReturn segment’s assets in a manner that will enable


16

this segment to outperform by a prudent but not heroicmargin the securities comprising the segment’sbenchmark. In pursuing this goal, the Committee andstaff will focus their energies on identifying outstandingexternal managers that can produce truly superiorreturns on a portion of the Total Return segment.Given the size of the Foundation’s aggregateendowment, the operative premise is that the percentageof the Total Return segment capable of producing trulysuperior returns will be small indeed, with most if notall of the return advantage gained through means otherthan strategic or tactical asset allocation.

D. Manager Selection

Manager Selection Criteria. In general, the moreimportant and desirable an attribute is (e.g., a well-defined and preferably innovative investmentphilosophy), the more difficult and time consuming itis to confirm its presence in a given money manager.Conversely, undesirable attributes (e.g., portfoliomanagers who are encumbered with excessiveadministrative or client servicing responsibilities) arerelatively easy to detect. This suggests that screeningcriteria should generally be applied in reverse order —negative screens first, positive screens later, after theselection universe has been reduced to a moremanageable number, as follows:

Step 1 Eliminate Managers Displaying Disqualifying Attributes• investment decisionmakers engaged primarily in

brokerage or financial planning (as distinct fromportfolio management)

• an inability to meet performance reporting deadlines• criminal convictions or sanctions by the SEC or

other federal or state regulatory agencies

Step 2 Eliminate Managers Displaying Too Many UndesirableAttributes• a high degree of personnel turnover• insufficiently trained administrative personnel• insufficiently robust investment accounting

systems• investment decisionmakers that are unduly

burdened with administrative tasks• an unwillingness to specify asset size limits for

products or services that require such limits• glaring mismatch between the liquidity offered

clients and the inherent liquidity of the strategiesbeing pursued


Step 3 Eliminate Managers Displaying Too Few DesirableAttributes• money management is the firm’s sole (preferably)

or primary line of business• the firm’s decisionmakers are seasoned

professionals or the firm’s philosophy is unusuallyinnovative (preferably both)

• the firm is willing to employ performance-basedfees (with meaningful hurdles rates that inhibitundue enrichment) as an expression of confidencein its own abilities

• the firm complies fully with the performancestandards promulgated by AIMR

Step 4 Eliminate Managers Displaying Too Few ImportantAttributes• a well-defined investment philosophy that gives

the manager a discernible competitive advantagein the gathering or processing of investment data

• a verifiable record that the firm has faithfullyexecuted this philosophy over time

• a proven capacity to deliver uniform results to allclients to whose assets the philosophy is applied

• a reasonable amount of assets under managementto which this philosophy is applied

• satisfactory returns versus relevant benchmarkindices

• a proven capacity to adapt to changes in financialmarkets

• a proven willingness to invest adequately in itsown business (including technological resources)in light of such changes

• investment professionals who have strong personalincentives (both financial and psychological) toproduce satisfactory results for their clients

• strong if not proprietary deal flow *

* Asterisk denotes criterion that is germane primarily to managersinvesting in non-marketable assets.

Liquidity Mismatch. In recent years, many moneymanagers have sought to feather their own nests byforming hedge funds whose redemption windows popopen at intervals much shorter than a prudent investmenttime horizon for the strategies such managers employ.The Foundation is acutely sensitive to the potentialproblems that this mismatch creates and seeks to avoidmanagers that have veered down this dangerous path.6

6 The path is dangerous because the constituencies that most hedge fundsserve bear little resemblance to the constituency served by the archetypicalpool operators, i.e., life insurance companies. Life insurers can toleratea mismatch between their investment time horizon and the periodicity offorced sales of investments because the catalysts for such sales (i.e.,deaths) tend to be independent events, and the sales tend to be partial. Theopposite approximates the truth for many hedge funds: the conditions thatinduce some clients to yank their money typically induce most clients towithdraw, transforming a trickle of outflows into a torrent of redemptions.

17


Exceptions That Prove the Rule. The policy biasadvanced in the preceding paragraph should not beconstrued as a prohibition against longer-term lock-ups for vehicles investing primarily in marketablesecurities. The Foundation is fully prepared to incurlong-term lock-ups with respect to marketable securitiesmanagers, as long as the handcuffs being adorned areassociated with demonstrably superior rules ofengagement in other respects, e.g., with respect to fees,clawback provisions, permissible strategies, and thelike. There are many hedge funds investing exclusivelyin marketable securities that would have great difficultyimplementing their favored strategies and tactics iftheir capital bases were subject to annual or even bi-annual withdrawals.

Expected Alphas. The asset allocation guidelines setforth in §II are accompanied by real return expectationsfor each segment that are themselves the sum of twoindependent variables: (a) an expected real return forthe assets comprising the segment’s benchmark if heldon a passive or indexed basis over multi-decade timeperiods and (b) an assumed premium from activemanagement (net of fees and trading costs). Thesepremia or “alphas” are necessarily “best guesses,” andare necessarily expressed in the form of point estimatesrather than ranges. In practice, the higher the expectedreturn on an asset class or subclass, the larger the rangeof expected alphas, i.e., the greater the dispersionbetween the long-term returns produced by the best-performing managers and those produced by the worst-performing managers.

* * * * *

Part II. Investment Policies

A. Return Objectives

Primary Return Objective. The Foundation’sprimary return objective is to preserve the purchasingpower of the unitized value of its endowment assets,net of administration and investment managementcosts, over rolling five-year periods. This goal issynonymous with the pursuit of a time-weighted returnon endowment assets that is at least equal to an inflationrate appropriate to the Foundation (the CPI or anagreed-upon alternative) measured over rolling five-year periods.

Secondary Return Objective. The Foundation’ssecondary return objective is to enhance the purchasing


power of the unitized value of its endowment assetsover rolling five-year periods. This goal is synonymouswith the pursuit of a time-weighted return onendowment assets that exceeds an inflation rateappropriate to the Foundation plus 5% net of all costs,measured over rolling five-year periods.

Tertiary Return Objective. The Foundation’s tertiaryreturn objective is to avoid peak-to-trough declines inunitized endowment purchasing power exceeding 30%.

Underpinnings. The adoption of objectives expressedin unitized terms reflects an assumption that theendowment’s total market value is likely to shrink inaccordance with the Foundation’s overall financialgoal of potentially liquidating its assets. The adoptionof rolling five-year periods for assessment of resultsreflects a balancing of the Foundation’s targeted lifewith its board’s perceived tolerance for unexpectedlypoor results. The five-year period for assessing resultsis intended as a floor but not a ceiling: its adoptionindicates that the board ascribes little importance todeclines in unitized endowment purchasing power thathave persisted for less than five years, unless theyentail peak-to-trough declines exceeding 30%. The30% limit serves as a proxy for the Foundation’stolerance for changes in its financial condition that,however fleeting, would likely trigger fundamentalchanges in the scope and character of its grantmakingendeavors.

B. Liquidity

Spending Policy. As a publicly supported organization,the Foundation is not legally obliged to distribute astated percentage of its assets in any given year, nor isit encumbered by grant commitments that are excessivein relation to its total assets. At present, suchcommitments approximate 12% of the Foundation’stotal assets in present value terms. In recent years,distributions have approximated 6% of theendowment’s market value. The board stands preparedto distribute more than the constant dollar equivalentof this sum in any given year if compelling grantmakingopportunities arise and less if depressed financialmarkets or a paucity of attractive grantmakingopportunities dictates a lower spending rate.

Cash Flow Requirements. The Investment Committeeseeks to ensure that the endowment comprises deflation


18

and inflation hedges (as defined below) whoserespective market values each are at least equal to 18-month projections of worst-case (i.e., maximum) cashflow requirements.

Cash Flow Requirements Defined. When quantifyingcash flow requirements as that term is used immediatelyabove, the Committee shall add to projectedgrantmaking needs (as defined by the board) themaximum aggregate capital calls the Foundation mightreceive from external managers to which it has agreedto furnish fresh capital. In general, because theFoundation has access to other sources of liquiditysuch as bank lines of credit, there is no need for theendowment to maintain substantial balances of veryshort-term instruments (so-called cash equivalents).

Rationale for 18-Month Projections. The requirementthat the endowment’s hedging segments (deflationand inflation, as defined below) be sized in accordancewith 18-month cash flow requirements is founded onthe assumption that economic shocks of either type(deflationary or inflationary) could cause the price ofmost assets held by the Foundation to decline materially.For example, in a major deflation, it is assumed that theprice of all assets except those held specifically fordeflation hedging purposes would decline materially.There is no assurance that the losses in question will bereversed within 18 months; indeed, history teachesthat major shocks of the sort being hedged againstalmost always depress asset prices for more prolongedperiods. The size of each hedging segment is peggedto 18-month cash flow needs, not because the disastersbeing “insured” against tend to be so short-lived, butbecause 18 months is deemed an adequate time periodfor the Foundation’s board and staff to adjustgrantmaking and perhaps other cash flows (e.g., thoseassociated with private investment activities) to suitchanging external circumstances. In an ideal world,the Foundation would take out insurance against theforced sale of assets at depressed prices underdeflationary or inflationary conditions lasting longerthan 18 months, but the assumed opportunity costs ofmaintaining hedges larger than those specified in theaccompanying exhibit (12.5% for each hedge or 25%in total) are deemed excessive. Moreover, under trulyextreme conditions (massive deflation or inflation), itis reasonable to assume that assets held to hedgeagainst such a condition would actually appreciate invalue, thereby obviating for periods longer than 18months the forced sale of other assets trading at verydepressed prices.

Privately Traded Assets. The Foundation may investwithout limitation in privately traded assets, exceptthat new agreements to purchase such assets shall notbe executed if the aggregate market value of all suchprivately traded assets exceeds 60% of the aggregatemarket value of the Foundation’s total assets.

C. Policy Portfolio

Purposes. The cornerstone for the management of theFoundation’s assets is its Policy Portfolio, which is setforth on page 23. The Policy Portfolio represents thelowest-cost asset mix that, in the InvestmentCommittee’s opinion, is likely to satisfy the objectivesspecified in §II.A and the constraints specified abovein §II.B.

Use and Abuse of Policy Portfolio. The Policy Portfoliois not intended to serve as a benchmark against whichthe short-term performance of the Foundation’s actualinvestments shall be measured. More specifically, theboard ascribes little importance to outperforming thePolicy Portfolio over rolling time periods shorter thanfive years. It recognizes that the Foundation’s actualreturns could lag the Policy Portfolio’s returns overany interim measurement period for two reasons.First, the Investment Committee has discretion to shiftfunds across segment boundaries (within prescribedlimits) to enhance returns or reduce risks. Whenexercising such discretion, the Committee recognizesthe extreme difficulty of making timely shifts acrosssegment boundaries. Indeed, as a general rule, theCommittee (and its delegates) refrain from makingmaterial or frequent shifts other than for rebalancingpurposes. Second, as noted below, the Committee hasdiscretion to deploy the capital allocated to eachsegment into holdings other than those comprisingeach segment’s benchmark. When exercising suchdiscretion, the Committee’s decisions shall employthe criteria for asset substitution set forth below. Absentinvestment opportunities appropriate to a fund segmentthat satisfy these criteria, the funds allocated to thesegment shall be deployed in a manner that will causethe segment’s performance to resemble as closely aspossible the performance of its benchmark.

Origins. The Policy Portfolio was established afterresearch and discussion involving board members,Investment Committee members, staff, and outsideexperts. Close attention was paid to the Foundation’sliquidity needs and perceived risk tolerance, as well as


19


7 In this context, the fallacy of composition warns trustees that if mostinvestors array their affairs in a manner which assumes that futureoutcomes will be bounded by historical extremes, then such extremes arevirtually certain to be eclipsed.

the projected behavior of the asset classes, subclasses,and strategies deemed worthy of consideration for theFoundation’s potential use.

Periodic Reviews. The Policy Portfolio is reviewedregularly by the Investment Committee and modifiedas needed in light of experience and changingcircumstances.

Conscious Offsets. The Policy Portfolio compriseshedging segments that, when constructed properly,could provide potentially offsetting total returns underextreme market scenarios. For example, if fundsallocated to the deflation-hedging segment areessentially “indexed” to the segment’s benchmark,then it is reasonable to suppose that in environments ofunexpectedly high inflation losses on such deflationhedges will essentially offset gains on the portfolio’sinflation hedges. The converse could occur indeflationary environments. Why does the PolicyPortfolio comprise these potentially offsetting hedges,in the normal proportions reflected in the accompanyingexhibit? The answer is rooted in three related factorsalluded to in §I of this document: (1) neither theFoundation nor any advisors available to it can forecastgeneral economic conditions more accurately andconsistently than “the market”; (2) although historyprovides some insight into the probable behavior ofvarious asset classes and subclasses under extremeeconomic conditions, the “fallacy of composition”makes it imprudent to assume that future outcomeswill necessarily be bounded by past extremes;7 and (3)the Foundation’s liquidity needs (i.e., grant-relatedoutflows) will not evaporate and could actually increaseunder extreme economic conditions. Taking all ofthese factors into account, the Investment Committeehas concluded that the Foundation should be adequatelyinsured against both disaster scenarios (prolongeddeflation or very high rates of unanticipated inflation),mindful that the opportunity costs of holding bothforms of insurance could exceed materially theopportunity costs of being less adequately insured.

Segment Return Objectives and Benchmarks. Thereturn objectives for each segment are based on ananalysis of capital market history, adjusted forvaluations and economic conditions at the time this

statement of policies was last revised. The benchmarksfor each segment reflect the Investment Committee’sbest judgment respecting the asset class or subclassthat is reasonably certain to achieve each segment’sreturn objective (net of fees) when bought and held onan indexed basis.

Total Return Segment Benchmark. This segment’sbenchmark is founded on the simplifying assumptionthat the segment’s allocation to private equity is a“given” for policy planning purposes. Accordingly,the segment’s benchmark is the weighted average of(1) actual private equity returns and (2) returns on adiversified global stock index with each return weightedin accordance with each subsegment’s proportionateweight. The operative premise underlying thisweighting scheme is that a material portion of the non-private equity assets held in the Total Return segmentwill comprise marketable securities, thus permittingthe Committee to rebalance the segment’s weightwithin the overall endowment without disturbingprivate equity holding. The choice of the MSCI AllCountry World Free Index as an appropriate benchmarkfor the non-private equity Total Return subsegmentcreates a rebuttable presumption that the Foundationwill maintain at least some exposure at all times to eachof this Index’s three generic parts: publicly traded USstocks (49% of the Index at year-end 1998), publiclytraded developed market foreign stocks (47% at year-end 1998), and publicly traded emerging market foreignstocks (4% at year-end 1998). The presumption isrebuttable because the actual allocation to each ofthese types of investments could approach zero ifsufficient opportunities arise to engage in AssetSubstitution (as defined immediately below).

Capitalization-Weighted Benchmarks. TheInvestment Committee recognizes fully the dangersinherent in basing actual portfolio choices on observedcapitalization weights (i.e., market weights) to theexclusion of other considerations. For example, it isevident in hindsight that institutions that havemaintained MSCI-like allocations to Japanese stocksover the last decade have paid a very heavy price fortheir aversion to benchmark risk (defined as deviationsfrom MSCI’s ubiquitous capitalization-weighted stockindexes). The Committee is mindful that someinstitutions have adopted policy portfolios comprisingnon-capitalization-weighted indices (e.g., GDP-basedapproaches), and it remains open-minded about thepossibility of adopting such approaches in due course.



20

8 As used throughout this document, “asset” is defined in the broadestpossible sense and includes securities or properties held directly, throughmanaged accounts, or through commingled vehicles (including fund-of-funds).

However, before adopting such approaches or evenagreeing to consider doing so, the Committee mustassess soberly whether the expected benefits of suchunconventional approaches are likely to exceed theexpected costs. These costs include the materialamounts of Committee and staff time that wouldassumedly accompany an effort to jettisoncapitalization-weighted benchmarks in favor ofalternate approaches; the upfront and ongoing hasslesassociated with selecting and monitoring externalmanagers willing to be measured against highlyunconventional benchmarks; and the stresses and strainsassociated with staff compensation schemes that linkquantitatively based bonuses (to the extent these arerewarded to key investment professionals) tooutperformance of benchmarks comprising unfamiliaror esoteric sub-benchmarks.

Asset Substitution. The Policy Portfolio is foundedon the assumption that the Foundation’s InvestmentCommittee will exercise judiciously its discretion toallocate assets within segments in a manner that willcause each segment’s actual returns to diverge favorablyfrom the returns of its specified benchmark.

Criteria. Committee decisions to employ assets8 otherthan those comprising each segment’s benchmarkshall be based on a careful assessment of the followingvariables: expected returns, net of fees, expenses and,with respect to leveraged investments, financing costs;a substitute investment’s capacity to reduce overallfund volatility; and the ease with which such assets canbe accounted for and traded. The weight accordedeach of these variables shall differ depending on thesegment in which an asset would be held.

Total Return Substitutes. Substitute assets held withinthe Total Return segment must have the potential tooutperform the securities comprising the segment’sbenchmark over a rolling five-year time period (orlonger periods in the case of investments entailinglonger-term lock-ups). As a general rule, substituteassets held within the Total Return segment are notexpected to contribute to the reduction of overall fundvolatility: under worst case conditions for the securitiescomprising the Total Return segment’s benchmark,

assets held within it are expected to decline materiallyin price rendering their immediate sale an unattractivemeans of meeting external cash flow requirements.

Hedging Substitutes. Substitute assets held primarilyfor hedging purposes (deflation or inflation) must havethe potential to outperform the securities comprisingthe relevant “naive hedge” over a rolling five-yeartime period (or longer periods in the case of investmentsentailing longer-term lock-ups). “Naive hedge” asused here means either 10-year US Treasury notes heldon a constant maturity basis (in the case of deflationhedges) or 10-year US Treasury Inflation-ProtectedSecurities (in the case of inflation hedges). As ageneral rule, substitute assets held for hedging purposesare not expected to produce long-term returnscommensurate with those produced by the securitiescomprising the Total Return segment’s benchmark.Assets that have the potential to produce such highreturns and to simultaneously reduce total fund volatilityduring either of the disaster scenarios being hedgedagainst (deflation or unexpectedly high inflation) shouldbe assigned to the Total Return segment.

Leverage. Under normal circumstances the Foundationwill not engage in borrowing for purposes of enhancingreturns. However, the Investment Committee hasdiscretion to purchase assets using borrowed money,provided that such debt financing does not exceed 5%of the total endowment’s market value when any suchpurchases are consummated nor 10% of totalendowment assets at any subsequent valuation date.

Alpha Estimates. When adding substitute assets toeach segment, staff shall propose and the InvestmentCommittee shall approve or modify estimates of eachprospective investment’s expected real return over itsexpected holding period defined three ways: (1) worstcase, (2) best case, and (3) best point estimate (i.e.,most likely outcome). The purpose of specifyingexpected returns in this manner is to ensure thatsubstitute assets embody an acceptable trade-offbetween risk and expected return. In the case ofactively managed holdings and excepting those caseswhere active excess returns (“alphas”) areindistinguishable from indexed or passive returns (e.g.,private equity), staff shall specify each of these twosources of total return under scenario (3) [i.e., bestpoint estimate], as well as the expected volatility ofsuch “alphas.” The purpose of specifying the expectedvolatility of holding-specific “alphas” is to ensure that


21


the information ratio of each actively managedinvestment (i.e., expected excess return divided by thestandard deviation of that return) falls within reasonablebounds.

Alpha Correlations. The Committee recognizes theextreme difficulty of forecasting correlations amongactive managers’ excess returns, especially under theworst-case (“macro shock”) conditions that makescenario testing a useful exercise in theory.Accordingly, the Committee encourages staff to redirecttime that would otherwise be spent forecasting suchcorrelations to ferreting out managers whose expectedalphas are abnormally skewed to the upside.

Portable Alpha. The Policy Portfolio reflects tworelated and important assumptions with respect to thedeliberate overweighting of any of the Portfolio’ssegments: first, the types of securities normallyrepresented in a segment will be overweighted (relativeto the segment’s normal weight) only when theInvestment Committee concludes that the excess returnopportunities presented by such securities are trulycompelling; second, the economic impact of fundingdecisions that would otherwise cause segment weightsto fall outside the ranges specified in the accompanyingguidelines shall be neutralized through the use ofderivative securities. As a general rule, the InvestmentCommittee will resist proposals to capture incrementalexcess returns through the overweighting of the TotalReturn segment relative to its rebalancing threshold of80%. This bias is founded on the assumption that whenTotal Return allocations approximate 80%, theendowment is likely already to have material exposureto the active strategies that would be funded with freshcapital. In light of the inherent conservatism of thebenchmarks adopted for the Hedging segments, nosuch bias exists with respect to the potentialoverweighting of these segments for alpha-capturepurposes provided that overweights falling outside theaccompanying rebalancing thresholds are neutralizedvia derivative securities or other means.

Asset-Specific Benchmarks. The primary objectiveof each and every asset held by the Foundation is toproduce returns at least equal to the benchmark for thesegment in which the asset is held. As previouslynoted, achievement of this objective shall be measuredover the longer of rolling five-year periods or therelevant lock-up period for investments entailinglonger-term lock-ups. As a means of assessing the

interim progress of substitute assets (i.e., theFoundation’s success in allocating funds to securitiesor strategies other than those reflected in each segment’sbenchmark), staff shall propose and the InvestmentCommittee shall approve the specification of a relevantmarket index for assessing each holding’s interimperformance. In some cases (e.g., with respect to USTreasuries held in the Deflation Hedge segment), themost relevant market index will be identical to thebenchmark for the segment to which the holding hasbeen assigned. In other cases (e.g., a managed accountof foreign stocks held within the Total Return segment),the chosen index will differ materially from thebenchmark for the segment in question. The morematerial the difference between the market indexdeemed most relevant to a specific holding and thebenchmark for the segment in which it is held, the moreconviction the Committee must have that the asset orstrategy in question can achieve the segment’s overallreturn objective taking the passive or indexed exposurethat it entails (if any) plus potential returns from activemanagement (if any) into account. The qualifier “ifany” is used in the preceding to underscore the fact thatsome assets that the Foundation might elect to holdentail passive exposure only (e.g., bond or stock indexfunds) while others are properly regarded as pure alphagenerators (e.g., certain forms of either fixed incomearbitrage or private equity investing).

D. Implementation

Staff Responsibilities. The Chief Investment Officer(CIO) and his staff are responsible for day-to-daymanagement of the endowment, for promoting theInvestment Committee’s discussion of asset allocationand manager selection opportunities and perils, forimplementing decisions made by the InvestmentCommittee, and for rebalancing the endowment asneeded to ensure compliance with the asset allocationranges set forth above.

Manager Hiring and Firing. The CIO has discretionto hire external managers provided that he/she providesto each member of the Investment Committee a duediligence memo respecting each prospective managerat least 15 business days in advance of the manager’sinitial funding. This 15-day notice requirement isintended as a floor rather than a ceiling, the assumptionbeing that staff and the Committee will conduct anongoing discussion of promising managers, with the



22

Committee approving informally the hiring ofdesignated managers prior to the commencement ofhighly time-intensive due diligence tasks (e.g., thepreparation of formal memos or the performing ofdetailed tax and legal analyses). The CIO also hasdiscretion to terminate external managers providedthat he/she provides to each member of the InvestmentCommittee written notice of the planned terminationat least five business days in advance of the termination.

Staff Compensation Plan. The Investment Committeebelieves that the Foundation’s long-term interests wouldbe well-served by adoption of compensationarrangements for the CIO and other senior investmentprofessionals comprising at least three parts: (1) basesalary, (2) a qualitatively determined bonus awardedand paid annually, (3) a quantitatively determinedbonus awarded annually but paid out in accordancewith “clawback” provisions designed to inhibit unduerisk-taking (e.g., year-end “gaming”). Thesearrangements are described in a separate document.

Investment Committee Responsibilities. TheCommittee shall oversee staff’s fulfillment of itsresponsibilities as specified above and shall also beresponsible for reviewing on an ongoing basis theperformance of the overall endowment and itsconstituent parts. Aided by staff, it shall also beresponsible for monitoring and exploiting opportunitiesto enhance returns through asset substitution as thattask is defined above.

Committee Composition. As noted in §I, in an effortto avoid “lowest-common-denominator” decision-making as well as excessive diffusion of responsibilityfor portfolio results, the Foundation’s InvestmentCommittee is small in comparison to peer group norms.

Current Size and Composition. At present, theCommittee comprises four voting members: theFoundation’s president; its chief investment officer;and two non-executive investment professionals, eachof whom has broad and extensive experiencesupervising institutional funds.

Future Additions. The Foundation is committed tokeeping the Investment Committee’s membership atfive individuals or fewer under all circumstances,including any staff members serving as voting membersof the Committee.

Decisionmaking Process. Given the Committee’srelatively small size and the extensive investmentexperience of each of its members, the InvestmentCommittee deems it wise to operate on a consensusbasis with each member enjoying a de facto veto overnon-delegable decisions (i.e., policy decisions that theCommittee rather than staff is authorized to make).

Rebalancing. Although last among the policy choicesthat collectively constitute asset allocation, this finalstep is arguably the most important. Perhaps the beststudy of rebalancing options (Arnott and Lovell, 1992)concludes that “disciplined rebalancing can boostreturns as much as a fairly large shift in the policy mixitself.” Interestingly, choosing a rational rebalancingrule is harder with respect to portfolios comprisingonly marketable securities than it is with respect toportfolios comprising both marketable and privatelytraded assets. It is harder because the presence ofprivately traded assets reduce dramatically the numberof times per year that a portfolio’s constituent parts canbe appraised for rebalancing purposes. Unfortunately,the simplicity associated with a smaller universe ofplausible rebalancing rules to choose among is morethan offset by the complexity of actually rebalancingportfolios comprising privately as well as publiclytraded assets.

Generic Rebalancing Rules. Mindful that theFoundation’s actual portfolio does indeed compriseilliquid assets, its choice of an appropriate rebalancingrule is essentially a choice between two simplealternatives:

Rebalancing to Norms. As soon as practicablefollowing the valuation of all of the Foundation’sassets, segment weights are restored to the targetsspecified in the accompanying exhibit; or

Rebalancing to Allowed Ranges. As soon as practicablefollowing the valuation of all of the Foundation’sassets, segment weights are adjusted, but only to theextent needed to move each weight back within itsspecified range.

Favored Rules. Which rule will work best in anygiven investment environment depends on a host ofinherently unforecastable variables, including thecorrelation of returns among the assets susceptible torebalancing, estimated trading costs, liquidity needs,and qualitative variables such as the Foundation’s


23


Policy Portfolio

Expected Returns Net of FeesExcess Return Real

Allocation Ranges Real from Asset TotalSegment / Eligible Assets Target Minimum Maximum Return Substitution* Return Reason(s) Held Benchmark

Total Return Assets 75% 60% 80% 6.3% 1.0% 7.3% Preserve and enhance Subsegment benchmarkspurchasing power weighted by normal %s

Existing Private per legal dictated by PI- 10.0% 0.0% 10.0% Enhance purchasing Actual PI returnsInvestments agreements related cash flows power

Other Total Return Assets 75% less PIs 5.0% 1.5% 6.5% Preserve and enhance MSCI All Country World• Developed Market Equities purchasing power Free Index adjusted for• Emerging Markets Equities dictated by asset substitution applicable withholding taxes• Private Equity Strategies criteria and liquidity constraints set forth

• Absolute Return Strategies in the accompanying document

• Other Assets and Strategies

Hedging Assets 25% 20% 40% 2.6% 0.5% 3.1% Avoid forced sale of Subsegment benchmarksTotal Return Assets weighted by normal %s

Inflation Hedges 10% 5% 15% 3.5% 0.5% 4.0% Avoid forced sale of 10-year US Treasury• Natural Resource-Related Assets Total Return Assets Inflation-Protected• Real Estate dictated by asset substitution during periods when Securities (TIPS) held on• Inflation-Linked Bonds criteria and liquidity constraints set forth unexpectedly high a constant maturity basis• Commodities in the accompanying document inflation causes such• Other Assets and Strategies assets to decline in price

Deflation Hedges 15% 10% 20% 2.0% 0.5% 2.5% Avoid forced sale of 10-year US Treasury note• US$ High Grade Bonds dictated by asset substitution criteria Total Return Assets held on a constant maturity• Non-US$ High Grade Bonds set forth in accompanying document during periods when basis

deflation causes suchassets to decline in price

Cash Equivalents 0% TBD** 15% 1.0% 0.0% 1.0% Liquidity buffer; source Merrill Lynch 182-dayof leverage to enhance Treasury Bill Index

returns in other segments

Total 100% 5.4% 0.9% 6.3% Constructed indexcomprising the assetsegment benchmarksspecified above weightedby the % allocation inthe “Target” column

* Asset substitution means the use of any managers, strategies, or tactics that could cause a segment’s or subsegment’s returns to deviate from the returns of its benchmark.** Minimum cash position could be negative, subject to discussions of appropriate leverage ratios.

image in the eyes of valued external managers. As ageneral rule, the first approach is potentially the mostprofitable provided that trading costs are controlledeffectively, and this is the approach favored by theFoundation when rebalancing capital across segments.Because rebalancing presupposes updated valuationson all of the Foundation’s investments, some of whichcannot be repriced more frequently than annually(e.g., selected private equities), actual segment weightsare unlikely to be adjusted back to specified targetsmore than once every 12 months. Intra-segmentrebalancing is dictated less by either of the rebalancingrules outlined above than by the asset substitutioncriteria set forth in §II.C.

MESSAGE IN A BOTTLE concluded

Rebalancing in Extreme Market Environments. Therebalancing rules employed by the Foundation undernormal market conditions must necessarily be modifiedin extreme market environments, which are defined asenvironments in which one of the three segmentsdescribed in the accompanying exhibit outperformsmaterially the other segments. In such environments,the Foundation will rely heavily on the sale ofappreciating assets to meet grantmaking needs, therebyreducing if not eliminating the need to liquidate otherassets at depressed prices.

!


24

THE INVESTMENT FUND FOR FOUNDATIONS590 Peter Jefferson Parkway, Suite 250

Charlottesville, Virginia 22911

Phone: 434-817-8200Fax: 434-817-8231Website: http://www.tiff.org

Electronic mail inquiries:Services offered by the cooperative: [email protected] account data: [email protected] selection procedures: [email protected]

Documents

Message in a Bottle - TIFF