DFYIF~MAY/JUNE 1Q95

Stephen C. Cetcheth is professor of economics at Ohio State University. The author thanks Margaret Mory McConnell for able researchassistance, ond Allen Berger, Ben Bernanke, Anil Koshyap, Nelson Mark, Alan Viard and the participants at the conference for comments andsuggestions. The author also expresses gratitude to the National Stience Foundation and the Federal Reserve Bank of Cleveland for financialand research support.

DistinguishingTheories of theMonetaryTransmissionMechanismStephen G. Cecchetti

yai raditional studies of monetary policys impact on the real economy haveI focused on its aggregate effects.

Beginning with Friedman and Schwartz(1963), modern empirical research in mone-tary economics emphasizes the ability of pol-icy to stabilize the macroeconomy. But casualobservation suggests that business cycleshave distributional implications as well, Oneway of casting the debate over the relativeimportance of different channels of monetarypolicy transmission is to ask if these distrib-utional effects are sufficiently important towarrant close scrutiny

The point can be understood clearly byanalogy with business cycle research moregenerally Ifrecessions were characterized bya proportionate reduction of income acrossthe entire employed populationfor example,everyone worked 39 rather than 40 hoursper week for a few quartersthen economistswould pay substantially less attention tocycles. It is the allocation of the burden orbenefit of fluctuations, with some individualsfacing much larger costs than others, that isof concern, There are two ways for an econ-omist to address this problem. The first is toattempt to stabilize the aggregate economy,the traditional focus of policy-orientedmacroeconomics, The second is to ask whythe market does not provide some formof insurance.

The recent debate over the nature of themonetary transmission mechanism can be

thought of in similar terms. According tothe original textbook IS-LM view of moneychanges in policy are important only insofaras they affect aggregate outcomes. Only thefluctuation in total investment is importantsince policies only affect the required rate ofreturn on new investment projects, and so itis only the least profitable projects (economy-wide) that are no longer funded. But sincethe most profitable projects continue to beundertaken, there are no direct efficiencylosses associated with the distributionalaspects of the policy-induced interestrate increase,

In contrast, the lending view focuseson the distributional consequences of mone-tary pohcy actions. By emphasizing a combi-nation of capital market imperfections andportfolio balance effects based on imperfectasset substitutability, this alternative theorysuggests the possibility that the policys inci-dence may differ substantially across agentsin the economy Furthermore, the policysimpact has to do with characteristics of theindividuals that are unrelated to the inherentcreditworthiness of the investment projects.An entrepreneur may be deemed unworthyof credit simply because of a currently lownet worth, regardless of the social return tothe project being proposed. It is importantto understand whether the investmentdeclines created by monetary policy shiftshave these repercussions,n

In this essay, I examine how one mightdetermine whether the cross-sectional effectsof monetary policy are quantitatively impor-tant. My goal is to provide a critical evaluationof the major contributions to the literaturethus far. The discussion proceeds in threesteps. I start in the first section with adescription of a general framework thatencompasses all views of the transmissionmechanism as special cases, thereby high-hghting the distinctions. In the second section,I begin a review of the empirical evidencewith an assessment of how researchers typi-cally measure monetary policy shifts, Thefollowing two sections examine the methods

The francial accelerator, ir whichthe impact ea ianestmeet of smallinterest changes is mognified bybalance sheet effects, is (Iso animportant part of many discussionsof the lending niew.

FEDERAL RESERVE SANK OF St. LOUIS

83

llFYIE~MAY/JUNE I 995

2 See lngersall (1987) fore conplete descf ptian of this problem.

[allowing the treditnnal francialeconomics approach, I bane avoirhddiscussing demand and supplyexplicitly. Irstead, the onsetdemands ore derined Irom the ideaof arbitrage relaflnnships among nilof tIne assets.

These include the limited parficipa-tan models based er Lucas(1990). See the survey byFexrst(1993), aswell as the summary inChdstiann and lichenhoaml1992L

used for differentiating between the theories.Studies fall into two broad categoriesdepending on whether they use aggregate ordisaggregate data. The third section discussesthe aggregate data, while the fourth sectiondescribes the use of disaggregate data, Aconclusion follows.

MONfl4Ufl( POUcYT.TWBIRY/T; tj~yflT;T;rcVt:rcfl51(

4~(

One way of posing the fundamentalquestion associated with understanding themonetary transmission mechanism is to askhow seemingly trivial changes in the supplyof an outside asset can create large shifts inthe gross quantity of assets that are in zeronet supply I-low is it that small movementsin the monetary base (or nonborrowedreserves) translate into large changes indemand deposits, loans, bonds and othersecurities, thereby affecting aggregate invest-ment and output?

The various answers no this puzzle canbe understood within the framework origi-nally proposed by Brainard and Tobin (1963).Their paradigm emphasizes the effects ofmonetary policy on investor portfolios, andis easy to present using the insights fromFamas (1980) seminal paper on the relation-ship between financial intermediation andcentral banks.

Famas view of financial intermediariesis the limit of the current type of financialinnovation, because it involves the virtualelimination of banks as depository institu-tions. The setup focuses on an investorsportfolio problem in which an individualmust choose which assets to hold given thelevel of real wealth. Labeling the portfolioweight on asset i as w,, and total wealth asW then the holding of asset, ithe assetdemandis just X, = w1W

In general, the investor is dividingwealth among real assetsreal estate, equityand bondsand outside money Each assethas stochastic return, ~,, with expectation ~and the vector of asset returns,~,has acovariance structure F. Given autility func-tion, as well as a process for consumption, it is

possible to compute the utility maximizingportfolio weights. These will depend on themean and variance of the returns.~and F,the moments of the consumption process,call these p.r, and a vector of taste parametersthat I will label 4, and assume to be constants.The utility maximizing asset demands can beexpressed as = w*,(~,fl ~

This representation makes clear thanasset demands can change for two reasons.Changes in either the returns process (~,For macroeconomic quantities (~,W) willaffect the XIs,3

At the most abstract level, financialintennediaries exist to carry out two functions.First, they execute instructions to changeportfolio weights. That is, following a changein one or all of the stochastic processes drivingconsumption, wealth or returns, the interme-diary will adjust investors portfolios so thatthey continue to maximize utility In addition,if one investor wishes to transfer some wealthto another for some reason, the intermediarywill effect the transaction.

What is monetary policy in this stylizedsetup? For policy to even exist, some gov-ernment authority, such as a central bank,must be the monopoly supplier of a nominallydenominated asset that is imperfectly substi-tutable with all other assets. I will call thisasset outside money In the current envi-ronment, it is the monetary base. There is asubstantial literature on how the demand foroutside money arises endogenously in thecontext of the type of environment I havejust described.4 But in addition, as Famaemphasizes, there may be legal requirementsthat force agents to use this particular assetfor certain transactions. Reserve requirementsand the use of reserves for certain types ofhank clearings are examples.

Within this stylized setup, apolicy actionis a change in the nominal supply of outsidemoney For such a change to have anyeffects at all, (1) the central hank controlsthe supply of an asset that is both in demandand for which there is no perfect substitute,and (2) prices must fail to adjust fully andinstantaneously Otherwise, a change in thenominal quantity of outside money cannothave any impact on the real interest rate, andwill have no real effects. But, assuming that

FEDERAL REEERVI RANK OF ST. LOUIS

84

DF~IF~MAY/JUNE 995

the policymaker can change the real returnon the asset that is monopolistically supplied,investors portfolio weights must adjust inresponse to a policy change.

The view of financial intermediaries thatis implicit in this description serves to high-light the Brainard and Tobin (1963) insightthat monetary pohcy can be understood byfocusing solely on the endogenous responseof investor portfolios. Understanding thetransmission mechanism requires a charac-terization of how asset holdings change inresponse to policy actions.

Second, even though there need be nobanks as we know them, there will surely beintermediaries that perform the service ofmaking small business loans. The agencycosts and monitoring problems associatedwith this type of debt will still exist, and spe-cialists in evaluation will emerge. While theywill have such loans as assets, they mostlikely will not have bank deposits as liabihties.Such entities will be brokers, and the loanswill he bundled and securitized,

With this as background, it is nowpossible to sketch the two major views of themonetary transmission mechanism. Thereare a number of excellent surveys of thesetheories, including Bernanke (1993a),Gertler and Gilchrist (1993), Kashyapand Stein (l994a) and Hubbard (1995).As a result, I will be relatively brief inmy descriptions.

flflE MON.EY VIEVIThe first theory commonly labeled

the money view, is based on the notion thatreductions in the quantity of outside moneyraise real rates of return,5 This, in turn,reduces investment because fewer profitableprojects are available at higher required ratesof returnthis is a movenuent along a fixedmarginal efficiency of investment schedule.The less substitutable outside money isfor other assets, the larger the interestrate changes.

There is no real need to discuss banksin this context, In fact, there is no reasonto distinguish any of the other assets ininvestors portfolios. In terms of the simpleportfolio model, the money view implies

that the shift in the w~sfor all of the assetsexcluding outside money are equal.

An important imphcation of this tradi-tional model of the transmission mechanisminvolves the incidence of the investmentdecline. Since there are no externalities ormarket imperfections, it is only the leastsocially productive projects that go unfunded.The capital stock is marginally lower. But,given that a decline is going to occur, theallocation of the decline across sectors issocially efficient.

This theory actually points to a measureof money that is rarely studied, Most empiri-cal investigations of monetary policy trans-mission focus on M2, but the logic of theportfolio view suggests that the monetarybase is more appropriate. It is also worthpointing out that investigators have foundit extremely difficult to measure economicallysignificant responses of either fixed or inven-tory investment to changes in interest ratesthat are plausibly the result of policy shifts,In fact, most of the evidence that is interpretedas supporting the money view is actually evi-dence that fails to support the lending view.

THE LIEHDff NO ViEW:BALANCE ~.w~rr:._;EPECTS

The second theory of monetary trans-mission is the lending view,6 It has two parts,one that does not require introduction ofassets such as hank loans, and one that does.The first is sometimes referred to as the broadlending channel, or financial accelerator, andemphasizes the innpact of policy changes onthe balance sheets of borrowers. It hearssubstantial similarity to the mechanism oper-ating in the money view, because it involvesthe impact of changes in th~real interest rateon investment.

According to this view, there are creditmarket imperfections that make the calcula-tion of the marginal efficiency of investmentschedule more complex. Due to informationasymmetries and moral hazard problems, aswell as bankruptcy laws, the state of a firmsbalance sheet has implications for its abilityto obtain external finance. Policy-inducedincreases in interest rates (which are both realand nominal) can cause a deterioration in

lerminalogy has the potenflnl tocreate confusion here. Ihave cho-sen the troditianal term for thistextbook IS-tM ar narrow moneynew. Ida not mean to imply thatthis is the monetaist view af thetransmission mechanism.

follaw Kashyap and Steins11 994al terminology rather thanthe more common credit xiew toemphasize the importance of loansin the erarsmissian mechanism.lerranke and Gerfier (1989,19901 pravide the original thearetcal underpinnings for this view.

FEDERAL RESERVE SANK OF ST. LOUIS

85

DFYIIE~MAY/JUNE 1905

lernunke, Gertler and Gilchriso(1994) refer to this nsa financioloccebruro -sinceitcoases smallchanges in interest rates no havepotentially large effects on invest-ment and output.

ltmaybe particularly difficult ox dis-tagaish these effects from thosethat arise from oaryiag cyclicnlity ofdifferentfirms soles and profitability.

o See James (1987) for a discussionof the oniqueness of bark loans.

mx With nominal rigidity, a decrease inoutside money reduces the pricelevel slowly, and so the real returnto holding money increases. Thischannel of transmission requiresthat inoestors shift awayfrom loansin response.Koshyap and Stein (1994b) pointout that large banks car issue (Osin away that insulates their halence sheets from contractor indeposits, hat small barks cannot.So long assmall banks are animportant source xl funds far samehnnkdependert firms, there willsoIl be a honk lending channel. Inother wards, for honk lending to bean important part of the traesmis-sian mechanism, credit marketimpemfectons mest be impartaetfar honks.

the firmsnet worth, by both reducing expectedfuture sales and increasing the real value ofnominally denominated debt, With lower networth, thefirm is less creditworthy because ithas an increased incentive to misrepresentthe riskiness of potential projects. Asa result,potential lenders will increase the risk pre-mium they require when making a loan. Theasymmetry of information makes internalfinance of new investment projects cheaperthan external finance.

The balance sheet effects imply that theshape of the marginal efficiency of investmentcurve is itself a function of the debt-equityratio in the economy and can be affected bymonetary policy7 In terms of a simple text-book analysis, policy moves both the IS andthe LM curves. For a given change in therateof return on outside money (which may bethe riskless rate), a lender is less willing tofinance a given investment the more debt apotential borrower has. This points to twoclear distinctions between the money andthe lending viewsthe latter stresses boththe distributional impact of monetary policyand explains how seemingly small changesin interest rates can have a large impact oninvestment (the financial accelerator).

Returning to the portfolio choice model,the presence of credit market imperfectionsmeans that policy affects the covariancestructure of asset returns, As a result, thew7s will shift differentially in response tomonetary tightening as the perceived riski-ness of debt issued by firms with currentlyhigh debt-equity ratios will increase relativeto that of others.8

The second mechanism articulated byproponents of the lending channel can bedescribed by dividing the other assets ininvestors portfolios into at least three cate-gories: outside money, loans and all theothers. Next, assume that there are firms forwhich loans are the only source of externalfundssome firms cannot issue securities.0

Depending on the solution to the portfolioallocation problem, a policy action maydirectly change both the interest rate and

the quantity of loans, It is not necessary tohave a specific institutional framework inmind to understand this, Instead, it occurswhenever loans and outside money arecomplements in investor portfolios; that is,whenever the portfolio weight on loans is anegative function of the return on outsidemoney for given means and covariances ofother asset returns.no

The argument has two clear parts. First,there are borrowers who cannot financenew projects except through loans, andsecond, policy changes have a direct effecton loan supply Consequently the mostimportant impact of a policy innovation iscross-sectional, as it affects the quantity ofloans to loan-dependent borrowers.

Most of the literature on the lending viewfocuses on the implications of this mechanismin a world in which banks are the only sourceof loans and whose habilities are largelyreservable deposits. In this case, areductionin the quantity of reserves forces a reductionin the level of deposits, which must bematched by a fall in loans. The resultingchange in the interest rate on outside moneywill depend on access to close bank depositsubstitutes. But the contraction in bank bal-ance sheets reduces the level of loans. Lowerlevels of bank loans will only have an impacton the real economy insofar as there arefirms without an alternative source ofinvestment funds.

As a theoretical matter, it is not necessaryto focus narrowly on contemporary banks intrying to understand the different possibleways in which policy actions have real effects.As I have emphasized, bank responses tochanges in the quantity of reserves are justone mechanism that can lead to a comple-mentarity between outside money and loans.Aspointed out by Romer and Romer (1990),to the extent that there exist ready substitutesin bank portfohos for reservable deposits suchas CDs, this specific channel could he weakto nonexistent,mn But it remains a real possi-bility that the optimal response of investorsto a policy contraction would be to reducethe quantity of loans in their portfolios.

The portfolio choice model also helps tomake clear that the manner in which policyactions translate into loan changes need not

FEDERAL RESERVE RANK OF SY. LOUIS

86

II E~IF~MAY/JUNE 1995

be a result of loan rationing, although itmaynz As Stiglitz and Weiss (1981) originallypointed out, a form of rationing may arise inequilibrium as a consequence of adverseselection, But the presence of a lendingchannel does not require that there beborrowers willing to take on debt at the cur-rent price who are not given loans. It ariseswhen there are firms which do not haveequivalent alternative sources of investmentfunds and loans are imperfect substitutes ininvestors portfolios.

Obviously, the central bank can takeexplicit actions directed at controlling thequantity of loans. Again, lowering the levelof loans will have a differential impact thatdepends on access to financing substitutes.But the mechanism by which explicit creditcontrols influence the real economy is adifferent question.tm

54cr645 SSxc~xx5 54554 56.

;wo VV~Ts OTVET~.CONBiiX lOTIONS

Distinguishing between these two viewsis difficult because contractionary monetarypolicy actions have two consequences,regardless of the relative importance of themoney and lending mechanisms. It bothlowers current real wealth and changes theportfolio weights.nx

Assuming that there are real effects,contractionary actions will reduce futureoutput and lower current real wealth, reducingthe demand for all assets. In the context ofstandard discussions of the transmissionmechanism, this is thereduction in investmentdemand that arises from a cyclical downturn.8

The second effect of policy is to changethe mean and covariance of expected assetreturns, This changes the w55s. In the simplestcase in which there are two assets, outsidemoney and everything else, the increase inthe return on outside money will reduce thedemand for everything else. This is a reduc-tion in real investment.

The lending view implies that thechange in portfolio weights is more complexand in an important way There may besome combination of balance sheet and loansupply effects.

This immediately suggests that lookingat aggregates for evidence of the right degreeof imperfect substitutability or timing ofchanges may be very difficult, What seemspromising is to focus on the other distinctionbetween the two viewsthe lending viewsassumption that some firms are dependenton loans for financing.

In addition to differences stemmingfrom the relative importance of shifts in loandemand and loan supply the lending viewalso predicts cross-sectional differences arisingfrom balance sheet considerations, These arealso likely to be testable. In particular, itmay be possible to observe whether, giventhe quality of potential investment projects,firms with higher net worth are more likelyto obtain external funding. Again, the majorimplications are cross-sectional.

EMPIRICAL t

Before discussing any empirical exami-nation of the monetary transmission mecha-nism, two questions must be addressed.First, do nominal shocks in fact have realeffects? Unless monetary policy influencesthe real economy it seems pointless to studythe way in which policy changes work.Second, how can we measure monetarypolicy? In order to calculate the impact ofmonetary policy we need a quantitativemeasure that can reliably be associated withpolicy changes.

Here I take up each of these issues. Inthe following section, 1 will weigh the evi-dence on the real effects of money This isfollowed by a discussion of ways in whichrecent studies have attempted to identifymonetary shocks.

%C OTtL c-ntcCLil c~54t45444554~5r cr54544944 555/~~

HtYIF~MAY/JUNE 1995

56 The equinalent apen economyobservation is that in small openeconomies, enchorge rates move inresponse to changes in policy.

xx Boschen and Mills (19921 describea related technique.

~See Hamilton 11994) for a completedescriptor af the methodalogy.

~The enact measrres and sample fol-low those of Krshyap and Stein(1 994rl, who kindly supplied thedata.

numerous monetary regimes, they argue thatapparently exogenous monetary policy actionspreceded output movements.

Recent researchers use more sophisticatedstatistical tools to study the correlationsbetween money and income. This money-income causality literature is largely inconclu-sive, because it fails to establish convincinglyeither that money caused output or thereverse. In the end, the tests simply establishwhether measures of moneyforecast output,not whether there is causation. Given thatoutside moneythe monetary baseis lessthan 10 percdnt of the size of M2, it is notsurprising that economists find the simul-taneity problems inherent in the questiontoo daunting and give up.

Two pieces of evidence seem reasonablypersuasive in making the case that moneymatters. First, the Federal Reserve seems tohe able to change the federal funds rate vir-tually without warning. (I am not arguingthat this is necessarily a good idea, just thatit is possible.) In the very short run, thesenominal interest rate changes cannot beassociated with changes in inflationaryexpectations, and so they must representreal interest rate movements, Such realinterest rate changes almost surely have animpact on real resource allocations,ra

The second piece of evidence comesfrom the examination of the neutrahty ofmoney in Cecchetti (1986, 1987). In thosepapers, I establish that output growth issignificantly correlated with money growthat lags of up to 10 years! There are severalpossible interpretations of these findings,but they strongly suggest that monetaryshocks have something to do with aggregatereal fluctuations.

II E~I[t~MAY/JUNE 199$

monetary policy episodes.This series looks extremely noisy and

it is hard to see how it could represent policychanges. The 1979-82 period is the onlyone with large positive or negative values.Although it is surely the case that there areunanticipated policy changes both when theFederal Reserve acts and when it does not,one would expect small normal shocks withoccasional spikes. If decisions are really thisrandom, there is something fundamentallywrong with the policymaking apparatus.Furthermore, since the federal funds rateitself is the equilibrium price inthe reservesmarket, given technicalities of the way thatmonetary policy is actually carried out, themarket-determined level of the funds rate isnot a policy instrument.20

The second figure shows the response ofthe log of the CPI to a positive one percentagepoint innovation in the federal funds rate. Tounderstand how this is computed, begin bywriting the vector autoregression as

Aa)y, =~x

where AU.) is a matrix of polynomials in thelagoperator L (LYx = yr,), y1 is the vectorof variables used in the estimation, and C 15mean zero independent (but potentially het-eroskedastic) error. The first step is to esti-mate the reduced form version of equation 1by assuming that no contemporaneous vari-ables appear on the right-hand side of anyequations (A(0) = I), This results in an esti-mate A(L) alongwith an estimated covariancematrix for the coefficient estimatescallthis (2. The impulse response functionsare obtained by inverting the estimated lagpolynomial B(L) =

But the point estimate of the impulseresponse function is not really enough toallow us to reach solid conclusions, It is alsoimportant to construct confidence intervalsfor the estimates. There are two ways to dothis. The first involves the technique thathas been called Monte Carlo Integration.This is a Bayesian procedure that involvespresuming that the distribution of the vectorof errors in equation 1the csis i.i.d.normal.22 To avoid making such stringentassumptions, I choose to estimate confidencebands using an alternative technique grounded

Esfimated Innovations to the FederalFunds RatesPercentage points

34