Financial Structures, Bank Lending Rates and Transmission Mechanism

Financial Structure, Bank Lending Rates, and the Transmission Mechanism of Monetary
Policy
Author(s): Carlo Cottarelli and Angeliki Kourelis
Source: Staff Papers - International Monetary Fund, Vol. 41, No. 4 (Dec., 1994), pp. 587-623
Published by: Palgrave Macmillan Journals on behalf of the International Monetary Fund
Stable URL: http://www.jstor.org/stable/3867521 .
Accessed: 26/12/2010 06:06
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .
http://www.jstor.org/action/showPublisher?publisherCode=pal. .
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Palgrave Macmillan Journals and International Monetary Fund are collaborating with JSTOR to digitize,
preserve and extend access to Staff Papers - International Monetary Fund.
http://www.jstor.org
IMFStaffPapers
Vol. 41, No. 4 (December1994)
© 1994InternationalMonetaryFund
Financial Structure, Bank Lending

Rates, and the Transmission
Mechanism of Monetary Policy
CARLO COTTARELLI and ANGELIKI KOURELIS*
The stickiness of bank lending rates with respect to money market rates is
often regarded as an obstacle to the smooth transmission of monetary
policy impulses. Yet, no systematic measure of the different degree of lend-
ing ratestickiness across countries has been attempted. Thispaperprovides
such a measure. It also relates the different degree of lending rate stickiness
to structural features of the financial system, such as the existence of
barriers to competition, the degree of development of financial markets,
and the ownership structure of the banking system. Thus, the paper
provides further evidence on the relationship between structural finan-
cial policies and monetary policy, as well as on the relevance of credit
markets for the monetary policy transmission mechanism. The role of
administered discount rates in speeding up the adjustmentof lending rates
is also discussed. [JEL: E43, E44, E52, E58]
T of monetary policy hinges on a set of crucial

HE EFFECTIVENESS
structural parameters-not directly controlled by central banks-
that reflect economic agents' reactions to policy impulses from money
* CarloCottarelliis a
Deputy Division Chief in the Monetaryand Exchange
PolicyAnalysisDivisionof the MonetaryandExchangeAffairsDepartment.He
is a graduateof the Universityof Siena and holds a graduatedegree from the
London School of Economicsand PoliticalScience.
Angeleki Kourelisis a Ph.D. studentat the Universityof Pennsylvania.This
paperwas writtenduringher IMF summerintershipat the Monetaryand Ex-
changeAffairs Department.
This paper benefited from discussionswith LeonardoBartolini and Seyed
HosseinSamieiandfromcommentsfromWilliamAlexander,ErnestoFeldman,
Sunil Sharma, and the participantsin a seminar held at the Monetaryand
ExchangeAffairsDepartment.We thankthe Bankfor InternationalSettlements
for providingsome of the data used in this paper.
587
588 CARLO COTTARELLIand ANGELIKI KOURELIS
markets. These structuralparameters(such as the elasticities of the

demandand supplyof financialand real assetsto moneymarketinterest
rates) are affected by the structureof the financialsystem, that is, the
existenceand degree of developmentof financialmarkets,the degreeof
competitionin these markets,and the availabilityof foreignsources of
finance.While economictheoryhas recognizedthis relation,1empirical
evidence on the subject has mainlyfocused on the effect of structural
changes in financialmarketson the demandfor money (for example,
Tseng and Corker(1991)). An aspect that has been almost completely
disregarded2 is the relationbetween financialstructureand the speed of
the monetarypolicy transmissionprocess.Thispapertakes up this issue
by focusingon how the financialstructureaffectsthe degreeof stickiness
of banklendingrates,thatis, the speedat whichbanklendingratesadjust
to their long-run equilibriumvalue after a "shock" affecting money
marketrates.
Recent economic literaturehas stressed that banks are not neutral
"conveyors"of monetarypolicyimpulses(BernankeandBlinder(1988),
Bernankeand Gertler(1989),Bernanke(1993)).Consider,for example,
a monetarypolicy tighteningreflectedin an increasein money market
rates. Such a tighteningmay fail to contain aggregatedemand or ex-
changerate pressuresif financialintermediariesdo not promptlyadjust
theirlendingrates.3The reactionof financialintermediariesis, of course,
moreimportantin developingcountries,wherethe directfinancialchan-
nels betweenprimarylendersandborrowersarelimited,butit is farfrom
irrelevantin industrialcountries.It is, for example,notablethatbetween
Januaryand September1992, when most Europeancentralbankswere
strivingto defend the exchangerate mechanism(ERM) paritiesby rais-
ingmoneymarketrates,the differentialbetweenmoneymarketandbank
lendingrates increasedsubstantially(by 100 basispoints in Swedenand
the United Kingdom,200 basis points in Denmarkand Italy, and over
300 basispointsin Finlandand Norway).Thissuggeststhat lendingrates
did not fully adjustto the changesin money marketrates.
To analyzethe relationbetweenbanklendingratestickinessandfinan-
'See, for example, Modiglianiand Papademos (1980), Vanhoose (1985),

Kareken(1984), and Faig-Aumalle(1987)).
2An exceptionis Pelzman(1969).
3 It couldbe
arguedthatthe behaviorof the lendingratebecomeslessimportant
if the demandfor bankdepositsis sufficientlyelastic.An increasein treasurybill
rates will move deposits out of the bankingsystem, thus affectingaggregate
demand throughthe availabilityof credit, ratherthan throughits cost. This
argument,however,disregardsthe factthatin manycountriesbankshavea large
buffer of governmentpaper that can be sold to counter the effect of deposit
changes(Rodrigues(1993)).
AND BANKLENDINGRATES
FINANCIALSTRUCTURE 589
cial structure,we follow a simpleapproach.First,we measurethe speed

of adjustmentof bank lending rates in 31 industrialand developing
countries,by regressingthe lending rate on a distributedlag of money
marketrates. Thisway, we estimatethe effect on lendingratesof shocks
in money market rates, the so-called "multipliers,"when the shock
occurs,afterthreemonths,aftersix months,andin the long run. Second,
we explainthe cross-countrydifferencesin thesemultipliersby regressing
themon severalvariablesrelatedto the structureof the financialsystem,
suchas the degreeof concentrationin the bankingindustry,the existence
of constraintson capitalflows and barriersto entry,and the size and the
efficiencyof the moneymarket.4We also examinethe role of administra-
tivelyset discountratesas instrumentsthat "signal"changesin the stance
of monetarypolicy, and their relationto bank lendingrate stickiness.
The paperis organizedas follows. SectionI discussesseveralchannels
throughwhichthe financialstructurecan affectthe stickinessof lending
rates. Section II presentsthe model used in the empiricalanalysisand
discussessome econometricproblemsrelatedto its estimation.Section
III summarizesthe resultsof the time series regressionsused to measure
the degree of stickinessof banklendingrates, while SectionIV presents
estimatesof the cross-sectionequationexplainingthe differencesin the
degree of stickiness.Finally,SectionV summarizesthe mainfindingsof
the paper and drawssome policy conclusions.
I. The Stickinessof Bank LendingRates

and the FinancialStructure
Definition of Lending Rate Stickiness
The term "interestrate stickiness"has taken on two related, but dis-

tinct, meaningsin the bankingindustry.First,it has been usedto indicate
thatbankratesare relativelyinelasticwithrespectto shiftsin the demand
for bank loans and deposits. Second, it has been used to mean that, in
the presence of a change of money marketrates, bank rates change by
a smalleramountin the shortrun(short-runstickiness),andpossiblyalso
4 Thus, this paper is clearly related to the "bank structure-performance" liter-

ature(see Heggestad(1979)and Gilbert(1984)for surveys,and Short(1979)for
a cross-countryanalysis).However, this literaturehas focused on the relation
betweenfinancialstructureand the levelof bankratesor bankrate differentials.
Whilethe level is importantfor the efficientallocationof resources,the dynamic
propertiesof the lendingrate are more relevantfor the transmissionmechanism
of monetarypolicy.
590 and ANGELIKIKOURELIS
CARLOCOTTARELLI
in the long run (long-runstickiness).In this paperwe will refer mainly

to the second definition.More specifically,we will focus on the reasons
for the existenceof "short-term"stickiness,an aspectthat we will show
to be empiricallymore relevantthan its "long-run"equivalent.5
Money marketrates will be defined as rates on short-termfinancial
instrumentsthat are not administrativelycontrolledby the centralbank.
We focus on these rates, ratherthan on administeredshort-termrates
(suchas discountrates),becausemarket-determined ratesare less likely
to be subject to different forms of "attrition"(for example, political
pressures)that can delay their adjustment(see also Section II).
The Relevance of the Financial Structure
The term "financialstructure"will be used fairlybroadlyto include

such features as the degree of developmentof money and financial
markets; the degree of competition within the banking system, and
betweenbanksand other intermediaries(as affectedby both the regula-
tory environmentand the numberand size of intermediaries);the exis-
tence of constraintson capitalmovements;and the ownershipstructure
of the financialintermediaries.
The relationbetween these featuresand bank lendingrate stickiness
can be explainedin four different,albeit related, ways.
AdjustmentCostsand the Elasticityof the Demandfor Loans
The bankingindustry,like any industry,faces adjustmentcosts when
prices (that is, interest rates) change. The degree to which these costs
delay the adjustmentof lendingratesto changesin money marketrates
dependson the elasticityof the demandfor bank loans, which, in turn,
depends on the structureof the financialsystem.
This argumenthas been formalizedby Hannanand Berger(1991),on
the assumptionthatthe bankloan marketis characterizedby monopolis-
tic competition,that is, each bank faces a downward-slopingdemand
curvefor its loans. In this case, a profit-maximizingbank that does not
face adjustmentcosts will alwaysset the lendingrate at the point where
the marginalrevenueon loans is equal to an exogenouslygiven money
marketinterestrate (Klein (1971)). Thus, the lendingrate wouldfollow
moneymarketrateswithoutdelay.6In the presenceof fixed adjustment
5 As will be shown,in most countries
lendingratestend to adjustalmostfully
to money marketrate changesin the long run.
6However,in the monopolisticcompetitionmodel, the changein the lending
rateis not necessarilyequalto the changein the moneymarketrate(see Cottarelli
and Kourelis(1994), p. 5).
FINANCIAL STRUCTURE AND BANK LENDING RATES 591
costs, however,the lendingrate will changeonly if those costs are lower

than the costs of maintaininga nonequilibriumrate.
If the demandfor loans is linear, the costs of preservinga nonequi-
libriumrate are equal to 0.25g(Am)2,where Am is the change in the
money marketrate and g is the derivativeof the demandfor loans with
respectto the lendingrate (HannanandBerger(1989)).Thismeansthat
the greater the elasticity of demand for loans, the higher the cost of
keepinglendingrates out of equilibrium.If we introducea time dimen-
sion, the above argumentimplies that a bank will prefernot to change
its lending rates if the discountedflow of lost profits arising from a
nonequilibirumpositionexceeds the fixed costs of changingthose rates.
In incompletefinancialmarkets,demandelasticityis likelyto be lower
in the shortrunthanin the long runbecause,in the long run,even in thin
financialmarketsthere are alternativesourcesof financeto bank loans.
But the differencebetween the short- and long-runelasticitiesexplains
why lending rates are stickierin the short run. If, in fact, the elasticity
of demandincreasesover time, the cost of beingoutsidethe equilibrium
in each periodand the discountedvalue of the streamof lost profitsalso
rises over time. A bankwill decide to raiselendingratesonly when that
present value exceeds the fixed costs involvedin changingthem; if the
elasticityof demand is lower in the short run, the adjustmentwill be
delayed.
Thus, the relationbetweenlendingrate stickinessand financialstruc-
ture is straightforward,as the financialstructureclearlyinfluencesthe
elasticityof demandfor loans. The demandfor loans of each bankis less
elasticin marketsthat have fewer competitors,higherbarriersto entry,
or no alternativefinancesources(suchas otherfinancialintermediaries,
foreigncapitalmarkets,commercialpaper,or bankers'acceptancesmar-
kets). In such markets, lending rates may show a limited response to
changesin money marketrates in the short run.
AdjustmentCostsand UncertaintyaboutFutureMoneyMarketChanges
In the presenceof adjustmentcosts, bankswill not adjusttheirlending
rates if they perceive that the changesin money marketrates are only
temporary.The uncertaintyregardingthe natureof moneymarketfluc-
tuationsprovidesan additionallink between lendingrate stickinessand
financial structure. Interest rate movements in insufficientlyliquid
moneymarketswill be characterizedby a strongrandomcomponentand
will not adequatelytransmitmonetarypolicy impulses,as policy signals
willbe lost in the noise of randommovements.As a result,the adjustment
of lending rates will be slower.
Non-Profit-Maximizing Behavior
The conclusionthat bank lendingrates adjustpromptlyto changesin
money market rates is based on the hypothesisthat banks maximize
profit.However,theremaybe financialstructureconditionsunderwhich
this hypothesisdoes not hold. This may be the case, for example, in
bankingsystemsdominatedby state-ownedbanks,in whichlendingrate
adjustmentsmay be delayeddue to politicalpressuresor simple ineffi-
ciency. In general,bankswill reactmore promptlyto changesin money
marketconditionsif non-profit-maximizing behavioris penalizedby mar-
ket forces. If marketforces are weak (for example, becauseof barriers
to entry, absenceof competitionfrom nonbankintermediaries,or con-
straints on internationalcapital movements), inefficiencywill not be
penalized,which may result in lendingrate stickiness.
OligopolisticCompetitionModels
Price stickinesshas often been considereda feature of oligopolistic
markets,becauseof the unpredictableresponseof oligopolisticcompeti-
tors to price changes, and/or the fact that oligopolisticcollusion may
break down when prices are changed. While there is not a monotonic
relationbetween the degree of stickinessand the concentrationof the
bankingindustry,some stickinesswhenthe marketdeviatesfromperfect
competition, at least until a clear market leader emerges, can be ex-
plainedby this feature. It can also be arguedthat, in oligopolisticmar-
kets, the stickinesscan be reducedif the centralbank acts as a market
leader by signalingchanges in the stance of monetarypolicy through
changesin an administereddiscountrate, as the latterreducethe uncer-
tainty about competitors'responses. This argumenthas been used to
explainthe strongempiricalrelationbetweenthe discountrate andbank
lendingrates observedin many countries.
II. The Empirical Model
Model Presentation
In order to analyzethe relationbetween lending rate stickinessand
financialstructure,a measureof the degreeof stickinessin variouscoun-
triesis necessary.To obtainsucha measure,we beginwith the following
dynamicmodel for the lendingrate:
ii,t = i,O + Pi,lii,t-1 + Pi,2mi,t + * + i,n+2mi,t-n + 3i,n+2+lAdi,t
+ + Ui,t,
+ Pi,n+3+jAdi,t_j-i (1)
FINANCIALSTRUCTURE
AND BANKLENDINGRATES 593
whereii,t,mitanddi,,are, respectively,the lendingrate,the moneymarket

rate, and the discount rate for countryi at time t. The index i ranges
from1 to M, whereM is the numberof countriesincludedin the sample,
while the time index t rangesfrom 1 to Ti.7The first differenceoperator
is signifiedby A, ui,tis an errorterm, and the pis are parameterswhose
values vary across countries. Equation (1) reflects a fairly common
approachto the modeling of the lending rate. Its steady state form
(omittingthe errorterm) is:
ii = PO(1 - P1) + [(P2 + *' + Pn+2)/(1 - Pl)]m, (2)
whichis consistentwith the monopolisticcompetitionmodel relatingthe

loan rate to the money marketrate (that is, to the exogenouslygiven
marginalyieldof alternativebankassets,or to the marginalcostof funds).
The fact that no other variableis assumedto affect the lending rate in
the long run is of coursea simplification.In a monopolisticcompetition
model of the bankingmarket,the lendingrate shouldalso be influenced
by shiftsin the demandfor loans, as well as by changesin the perceived
riskinessof loans. These variableswere omitted in order to keep the
estimatedmodelsufficientlyconcise.The possibleomissionof some vari-
ables explainswhy the errorterm in equation(1) cannotbe assumedto
be seriallyuncorrelated.We do assume,however,thatuitis uncorrelated
acrosscountries.8
The dynamicspecificationreflectsa partialadjustmentmodelin which,
alongwith the laggeddependentvariable,the currentand severalof the
lagged values of the money market rate are included. In addition, a
polynomial distributedlag of the change in the discount rate is also
included.Thisreflectsthe hypothesis,discussedin SectionI, thatchanges
in the discountrate speed up the adjustmentof lendingrates, with no
effect on their long-runequilibriumvalue.
Given the cross-countrydifferencesin the 3sin (1), lendingrateswill
show a different degree of stickinessin response to shocks in money
marketrates. The followingprocedurewas followedto derivesummary
measuresof the degree of stickiness.Fromequation(1) we derivedsets
of "multipliers"reflectingthe adjustmentof the lendingrate duringthe
periodwhen the moneymarketrate changes(impactmultipliers),and at
7 Notice that the sample period varies across countries (see Section III).
8
This, of course,does not meanthatthe interestratesare uncorrelatedacross
countries, but, rather, that the cross-countrycorrelationof interest rates is
transmittedthroughmoney marketrates.
differenttimelags(interimmultipliers).Thesemultipliersare, in general,
deterministicnonlinearfunctionsof the Bs:
hi,e = 4([i), (3)
where hie is the value of the multiplierfor countryi aftert periods; )(.)
is a nonlinearfunction (see Appendix);and i is a vector of estimated
coefficientsfor countryi. We assumethat the value taken by the multi-
pliers depends on the structuralfeaturesof the financialsystem:
hi,e = ZiYe+ vi,e, (4)

where Zi is a K-element vector describingthe financial structureof
economy i and vi, is an error term uncorrelatedacross countries. In
matrixform equation (4) can be written, for differentlags, as:
ho = Zyo + vo; (5)

he = Zye + ve; and (6)
hL = ZYL + VL, (7)
where ho is a vector of impact multipliers(t = 0); he is a vector of
"interim"multipliersreflectingthe adjustmentof the lendingrate after
f periods;and hL is a vector of long-termmultipliersreflectingthe total
adjustmentof lendingrates (all these vectorshave M elements).Z is an
(M x K) matrixof structuralvariables,and the v vectors are (M x 1)
vectors of homoscedasticresiduals,which are assumedto be indepen-
dently distributednot only across countries,but also acrosstime lags.
The mainfocus of this paperis the estimationof the y vectorsdescrib-
ing the relation between the structuralvariablesand the h multipliers,
that is, our measure of lending rate stickiness. A two-step estimation
process was followed. In the first step (Section III), equation (1) was
estimated for 31 countries. Then, by filteringthe estimated P vectors
throughequation (3), an estimatefor the h vectorswas derived. In the
second step (Section IV), the estimatedvectorswere regressedagainst
the structuralvariablesincludedin Z.
Discussion of the Model
Before movingto the next section, some of the featuresof the above
empiricalmodel must be discussed.
Definitionof the Multipliers

The multipliersdefined above refer to the effect of a change in the
moneymarketratefora givendiscountrate.Wefocuson thesemultipliers
becausethe stickinessof banklendingratesemergesmore clearlyin the
absenceof discountrate changes. Indeed, as arguedabove, oligopolies
are expected to respondfairly quicklyto changesin the discountrate.
It could be argued that, from a policy perspective, the reaction of
lendingrates to both money marketand discountratesshouldbe exam-
ined, since they are both controlledby the monetaryauthorities.How-
ever, the discountrate is often not a marketrate, but is set administra-
tively. Unfortunately,administeredrates may themselvesshow a high
degree of stickiness, as they may be subject to more direct political
pressures,and often requirecomplex administrativeprocedures.Thus,
a transmissionmechanismcenteredon discountratechangesmaybe less
effective than a transmissionmechanismrelyingonly on money market
changes-hence the need to assess the stickinessof lendingrates in the
absence of changesin the discountrate.
The RelationBetweenthe 1 Coefficientsand the h Multipliers
In the above model, the multipliersh, ratherthan the 13coefficients,
are modeled as a linear functionof the structuralvariablesZ, because
as discussed in Section I, there is a relation between the structural
variablesand the size of the adjustmentat differentlags, whichis mea-
suredby h. One could be temptedto assumea directrelationbetween
the P3sand the Z matrixbut this would be inappropriate.Consider,for
example, the followingdistributedlag model (the i subscriptis omitted,
for simplicity):
it = PO + f2mt + P3mt-1 + P4mt-2 + ut. (8)
Suppose that the 13coefficientshad been modeled directlyas a linear

functionof the variableincludedin Z; for example:
33= Z4 + x. (9)
Based on the discussionin Section I, we expect that an increasein, say,
variable Zkwill lead to a faster adjustment, that is, to a larger multiplier
aftertwo periods.Thisrequiresa largersum 12 + 13withrespectto other
countriesbut it does not constrainthe value, or even the sign of the
coefficientof Zk in equation(9). An increasein Zkmaylead to a decrease
or an increasein 13, dependingon whether12 increasesby more or less
than 32 + P3. Since Zkaffects the sum of two coefficients, we cannot infer
the effect of Zk on one of the two.
The DynamicSpecificationof the Model

and the Two-StepEstimationProcedure
The relation between the p coefficientsand the h multipliersis, as
noted above, nonlinear,becauseof the inclusionof the laggeddependent
variable in equation (1). If equation (1) did not include the lagged
dependentvariable,equation(3) wouldbe linear.In this case, by substi-
tuting(3) into (1), the lendingrate couldbe expressedas a functionthat,
while nonlinearin the variables,wouldbe linearin the parameters,and
couldbe estimateddirectly.A two-stepestimationprocesswouldnot be
necessary.
The adopted specificationis preferablefor two reasons. First, a dy-
namicspecificationincludingthe lagged dependentvariableis typically
more parsimoniousthan the one based on distributedlags. Second, and
moreimportant,even if the relationbetweenlendingratesandstructural
variableswere linear in the parameters,its direct estimationwould be
extremelycumbersome.In fact, each structuralvariablewould appear,
on the right-handside, multipliedby severallags of the money market
rateandthe discountrate.Thus,for example,with 10structuralvariables
and 12 lags for the moneymarketand the discountrates, therewouldbe
as manyas 240 regressors,characterizedby a highdegreeof collinearity.
This would make any serious specificationsearchvirtuallyimpossible.
RelationbetweenMultipliersat DifferentLags
The parametersy in equations (5)-(7) are not independentacross
equations. This is intuitive because all the ys at differentlags (for an
arbitrarilylong lag length) are a functionof a limited numberof the P3
parametersin (1). However,the relationbetweenthe ys cannotbe easily
exploitedto improvethe efficiencyof the estimation,becauseit involves
not simplelinearconstraintson the ys butnonlinearconstraintson linear
combinationsof the ys (see Appendix).Therefore,the existenceof these
constraintswill be ignoredin the second step of the estimationprocess.
h Multipliersare Not Observedbut Estimated
If the h multiplierswere directlymeasured,the ordinary-least-squares
(OLS) estimatesof equations(5)-(7) would be unbiasedand efficient.
However,the h multipliersare estimatedfromequation(1). Therefore,
in orderto estimatethe y vectorsefficiently,equations(5)-(7) must be
estimatedafteradjustingfor heteroscedasticity,thatis, throughweighted
leastsquares(Saxonhouse(1976),(1977),CottarelliandKourelis(1994)).
The use of weightedleast squaresrequiresestimatingthe varianceof
the elementsof the h vectors. Suchan estimatecan be derivedeasilyfor
FINANCIALSTRUCTURE
the impactmultiplierho = 13o(whose varianceis estimateddirectlyfrom

equation(1)), but is more problematicfor the interimmultipliers,since
they are nonlinearfunctions of the P coefficients. Consequently,the
discussionin SectionIV will focus mainlyon the impactmultipliers,that
is, on the estimationof equation (5).
NonlinearRelationbetweenh and Z
Equations(5)-(7) postulatea linear relationbetween the multipliers
h andthe structuralvariablesZ. One problemwiththisassumptionis that
for certainvalues of the Z variables,the multiplierscould become neg-
ative (implyingthat the lendingrate declineswhen money marketrates
are raised).The standardsolutionto this problemwouldbe to impose a
nonlinearrelationbetweenh and Z, so that for any value of Z, h would
alwaysremainpositive. A simple way of doing so is to assumethat the
relationbetween h and Z is describedby a logistic function:
h = c/[l + exp(-Zy)]. (10)
This way h would be constrainedbetween 0 and c (a fixed parameter).
By takinglags, equation (10) could be linearized:
log(clh - 1) = -Zy (11)
This approachwould not be problematicif h were observed,but, as it
is not, the error term would enter equations (5)-(7) in a nonlinear
fashion. The originallinear formulationwas thereforemaintained.As
will be shown, this does not seem to create problemsin the estimation
of equation (5), as all fitted values remainedpositive.
III. Step One: Analysis of Stickiness
The Data
Havinglaid out the schemeof the empiricalmodel, we can proceedto

its estimation. The estimation of equation (1) for different countries
requiresmonthlyseries of lending rates, money marketrates, and dis-
count rates.9These data must be availablefor a sufficientlylong period
duringwhich lending rates were not administrativelycontrolledby the
9High frequencydata are needed because time aggregationmay bias the

estimates. Moreover, for the purpose of policy analysis, monthly lags are cer-
tainly more relevant than quarterly lags.
central bank10and no direct controlson the amountof credit were in

place.11This limitedthe samplesize to 31 countries,almostequallysplit
between developingand industrialcountries.It also limitedthe sample
period, sometimesto no more than two years.
Three types of lending rates were used: posted prime rates, posted
nonprimerates, and averagerates actuallychargedon bank loans. The
fact that these ratesmay show differentdynamicpropertieswith respect
to money marketrates was ignored in the first step of the estimation
process, but was taken into accountin the second step (see Section IV).
The data on money marketrates usuallyrefer to either treasurybill or
interbankrates. Discount rates refer to interestrates on variousforms
of last resort credit from the centralbank.
The stationarityof the above 93 series (three series for each of the 31
sample countries) was assessed using augmentedDickey-Fullertests.
Sincealmostall the serieswerefoundto be nonstationary,the modelwas
estimatednot only in levels but also in first differences,which in most
cases was sufficientto removethe nonstationarity.12 We will referthere-
fore to two sets of results:"Model 1" results (estimatesin levels), and
"Model 2" results (estimatesin first differences).
Estimation Results
Table 1 shows the estimatedmultipliersof changesin money market

rates at differenttime lags for both Model 1 and Model 2. (Detailed
resultsare presentedin Cottarelliand Kourelis(1994)). With reference
to Model 1, columns1-4 report,respectively,the impactmultiplier,the
multiplierafterthree monthsandthat aftersix months,andthe long-run
multiplier.The sameinformationfor Model2 is reportedin columns5-8.
The last two rowsof the table reportthe mean and the variationcoeffi-
10In orderto allowsomeinitialadjustmentof the

lendingrateto its equilibrium
level after the removal of interest rate ceilings, the sample periods used for
estimate(1) in countriesthatexperiencedceilingsstartedat leastsix monthsafter
their
n Ifremoval.
directcontrolsare in place, the relationbetweenlendingratesand money
marketrates is severed. This is because, in the absenceof creditrationing,the
lendingratewill be determinedby the intersectionof the demandfor bankloans
and the administrativelyfixed supplyof bank credit. In this case, a changein
money marketrates may not bringabout any changein the lendingrate. Thus,
if the stickinessof lendingrates were assessedduringperiodsof bindingdirect
controls,the degree of stickinesswouldprobablybe overestimated.
"2Theonly exceptionis Poland,for which, due to the limitedsamplesize, it
wasnotpossibleto estimatethe modelin firstdifferences(firstdifferences models
are likely to requirelonger distributedlags).
Table 1. Multipliers
(Effect on the Lending Rate of Changesin Money Marke
Model 1
Country Impact 3-month 6-month Long-run Impact
Australia 0.11 0.40 0.60 1.17
Belgium 0.21 0.61 0.81 1.03 0.21
Canada 0.76 0.93 1.00 1.06 0.78
Colombia 0.42 0.87 0.97 1.03 0.44
Denmark 0.07 0.25 0.38 0.71 0.15
Finland 0.13 0.20 0.27 0.60 0.13
Germany 0.38 0.67 0.83 1.04 0.37
Greece 0.40 0.74 1.05
Hungary 0.09 0.31 0.47 0.88 0.19
Iceland 0.61 1.04 1.07 1.08 0.61
Indonesia 0.19 0.59 0.84 1.21 0.20
Ireland 0.32 0.80 0.96 1.03 0.34
Israel 0.77 1.22 1.24 1.25 0.77
Italy 0.11 0.40 0.61 1.22 0.12
Jamaica 0.15 0.38 0.66 0.92 0.24
Table 1. Multipliers(concluded)
(Effect on the Lending Rate of Changes in Money Marke
Model 1
Country Impact 3-inonth 6-month Long-run Impact
Japan 0.06 0.19 0.25 0.75 0.03
Malaysia 0.16 0.29 0.39 0.91 0.13
Mexico 0.83 1.40 1.34 1.29 0.72
Netherlands 0.52 0.97 1.03 1.04 0.52
New Zealand 0.09 0.48 0.60 0.67 0.11
Philippines 0.27 0.75 0.81 0.87 0.24
Polanda 0.04 0.15 0.24 0.59 -
Portugal 0.28 0.77 0.97 1.12 0.47
Singapore 0.27 0.71 0.83 1.00 0.27
South Africa 0.61 0.79 0.88 0.99 0.73
Spain 0.35 0.80 0.98 1.12 0.36
Sri Lanka -0.22 0.28 0.30 -
Swaziland 0.48 0.52 0.54 0.57 0.54
United Kingdom 0.82 1.02 1.04 1.04 0.87
United States 0.32 0.69 0.85 0.97 0.41
Venezuela 0.38 1.03 1.30 1.48 0.24
Mean 0.32 0.64 0.77 0.97 0.33
Variationcoefficient 0.79 0.51 0.40 0.25 0.78
a
Model 2 was not estimatedfor Poland.
Table2. Correlation
BetweenMultipliers
(At differentlags)
Model 1 Model 2
ho h3 h6 hL ho h3 h6 hL
ho 1.00 0.89 0.77 0.46 1.00 0.80 0.70 0.63
h3 1.00 0.96 0.67 1.00 0.92 0.88
h6 1.00 0.80 1.00 0.93
hL 1.00 1.00
cient (that is, the ratio between standarddeviationand mean) of each

column. The followingfeaturesare notable.
ResultsRobust withRespectto Model Specification
Models 1 and 2 yield very similarmeasuresof the multipliers.The
correlationcoefficientbetweenthe impactmultipliersof Model 1 and of
Model 2 (that is, betweenthe first and fifth columnsof Table 1) is 0.97.
The correlationcoefficientdeclines slowly at longer lags: it is 0.91 and
0.86, respectively,for three-andsix-monthmultipliersbut remainsfairly
high(0.74) even for the long-runmultipliers.Thus, the resultsare robust
with respect to differentmodel specifications.
Degree of StickinessHigh on Average
The degree of stickinessis, on average,relativelyhigh. While, in the
long run, the lendingrate seems to adjustfullyto the moneymarketrate
(the long-runmultiplieris, on average,0.97, and it fallswithinthe range
of 0.75-1.25 in three fourthsof all cases), the impactmultiplieris only
one-thirdof the long-runmultiplier.13 Broadlyspeaking,thisimpliesthat
in order to increaselendingrates by 100 basis points duringthe month
of the money marketrate shock, the money marketrate must be raised
by 300 basis points. On average, after three months and six months,
respectively, about one-third and one-fourth of the adjustment re-
mains to be completed. Moreover, the ordering of the countries by
degree of stickinessis not affectedmuch by the lag at whichmultipliers
are measured(see Table 2).
StrongCross-CountryDifferences,Particularlyat ShortLags
There is much cross-countryvariationaroundthese averagevalues,
particularlyfor shorterlags. The standarderroris about80 percentof the
meanfor the impactmultiplierbutdropsto 50 percentafterthreemonths
and to 25 percent in the long run. Thus, countriesseem to differ more
13 Given the
similarityof the resultsof the two models, we will commentonly
on the Model 1 estimates.
602 CARLOCOITARELLIandANGELIKIKOURELIS
in the short than in the long run. This has two implications. First, it sug-
gests that the effect of different financial structures can be better assessed
by looking at short lags, rather than at long lags, a feature that will also
be evident from the results of Section IV. Second, this result is consistent
with the fact that the strong short-run differences are due to adjustment
costs or "inefficiencies," rather than long-run differences in loan demand
elasticities. The effect of these adjustment costs and inefficiencies tends
to fade away in the long run.
The differences among impact multipliers across countries cannot
easily be related to the degree of development of the economy. Focusing
on the impact multipliers, the subsample of countries represented by
higher-than-average performers (that is, those with an impact coefficient
higher than 0.32) is almost equally split between industrial and develop-
ing countries. The same is true for below-average performers. Clearly,
an explanation of the cross-country differences must go beyond a simple
consideration of the degree of overall development of the economy.
Relevance of Discount Rate Changes
The effect of discount rate changes on lending rates for the countries
in which such a variable was significant is reported in Table 3. The
discount rate appears to be a powerful instrument for speeding up the
adjustment of the lending rate to money market shocks. The discount
rate is significant in about one-half of the sample countries. Among these
Table 3. Effectof Changesin the DiscountRate
Model 1 Model 2
Country Impact 3 months 6 months Impact 3 months 6 months
Australia 0.20 0.14 0.11 0.27 -
Belgium 0.68 0.35 0.17 0.58 0.17
Denmark 1.25 0.91 0.66 1.00 0.34
Finland 0.45 0.38 0.32 0.41 0.19-
Germany 0.23 0.12 0.07 0.17 -
Iceland 0.25 0.02 0.01
Ireland 0.36 0.23 0.07 0.25 -
Italy 0.63 0.62 0.46 0.51 0.19 0.06
Japan 0.07 0.29 0.26 0.09 0.33 0.18
Netherlands 0.69 0.09 0.01 0.51 -
Poland 1.03 0.83 0.66 1.03 0.83 0.66
South Africa 0.19 0.10 0.06 0.09 -
Sri Lanka 0.15 0.20 0.05 0.19 --
Swaziland 0.32 0.19 0.11 0.23-
United States 0.49 0.21 0.09 0.29 -
Mean
Mean 0.47 0.31 0.21 0.40 0.34 0.30
countries,the averageimpactmultiplierof a changein the discountrate

is 0.47, and in some cases it is as high as 100 basis points. When the
discountrate is changed, the percentagemultiplierrises to 89 percent
from 26 percent, a threefold increase in the speed of adjustmentof
lending rates.
One importantfeature of the countriesin which the discountrate is
significantmust be noted. In the absence of a discount rate change,
lendingratesin those countriesshowa belowaverageresponseto money
market changes. Their average impact multiplieris 0.26, against an
averageof 0.36 for the othercountries,14 whichsuggeststhatthe stickiness
of lendingratesandthe effectivenessof the discountratemaybe related.
If a relationexists, it could be interpretedin two ways. On the one hand,
it could be arguedthat, in the presenceof a weak financialstructureand
sticky lending rates, monetary authoritieshave to rely on publicized
discountrate changesto spurthe bankingsystem. On the other hand, it
couldbe claimedthatin countrieswherethe centralbankhascustomarily
reliedon discountratesignals,bankshavebecome "addicted"to the use
of this instrument,to the extentthatlendingratesare not changedunless
the discountrate also changes.
Both these interpretationsimplya negativestatisticalrelationbetween
the impactmultipliersof money marketchangesand those of discount
rate changes. The first interpretation,however, also implies that the
stickinessof lendingratescanbe explainedpurelyby lookingat structural
variables.If the financialstructureis responsiblefor both the stickiness
of lending rates and the use of the discountrate as a monetarypolicy
signal,it shouldbe possibleto estimatea reducedformequationin which
the money market multipliersare uniquely related to the structural
variables.However,thiswouldnot be possibleif, in additionto the effect
of the financialstructure,the use of the discountrate as a policy signal
furtherreducesthe multipliers.In this case, a negativedummyequal to
1 when the discountrate is used as a "policysignaling"deviceshouldbe
significant,and with negative sign, in the regressionof Section IV.
IV. Step Two:Determinants

of the Stickinessof LendingRates
We now focus on the factorsexplainingthe cross-countrydifferences
in the stickinessof lending rates.
14
Forthe countriesin whichthe discountrate is significant,thereis a negative
correlationbetweenthe size of the impactmultiplierof moneymarketratesand
those of the discountrate.
CARLOCOTTARELLI
Structural Variables
Steptwo-the estimationof the relationbetweenmultipliersandstruc-

turalvariables-requires the identificationand measurementof the lat-
ter. Based on the discussion in Section I, four groups of structural
variables(reflectingthe degreeof bankcompetition,the extentof money
marketdevelopmentandthe opennessof the economy,the public/private
natureof the bankingsystem, and the overalldegree of developmentof
the financialsystem)havebeen singledout. In addition,it was necessary
to control for some additionalfactors affecting the dynamicsof the
lendingrates, suchas the differentinflationaryenvironment,the type of
the lendingrate series used in step one, and the use of the discountrate
as policy signal.
Before proceeding,two caveatsare necessary.First, while the range
of structuralvariablesincluded is large (given the limited number of
observationsavailable),it may not be exhaustive. Probablythe most
importantomission, due to insufficientdata availability,is the absence
of variablesreflecting the barriersto competitionbetween bank and
nonbankfinancialintermediaries.This will have to be borne in mind
when interpretingthe results.15
Second, it must be stressedthat, while the followingvariablescan be
definedas "structural,"they arenot fixedovertime. Thisdoes not create
a problemin the majorityof cases in whichno majorstructuralchange
(such as the removalof barriersto entry) occurredin the period over
whichthe multipliersweremeasured.Thenthe structuralvariablescould
be measuredat any period of time, and, indeed, were sometimesbased
on a single annualobservation.However, when structuralchangesoc-
curredor wheneverinformationon the structuralvariableswas available
overtime, the structuralvariableswerecomputedby usingaveragevalues
over the sample period.16
Competitionwithinthe BankingSystem
As in most studies of the relation between banking structureand
performance,the degree of competitionwithinthe bankingsystemwas
proxiedby variablesmeasuringthe degreeof concentrationof the bank-
ing system,such as the marketshareof the largestfive banks(MARSH)
15Omittedvariables resultsif theyarecorrelated
canbiasthe regression with
the includedvariables.Thismaybe the case, forexample,becauseregulatorsmay
constrainboththe competitionwithinthe bankingsystemandthatbetweenbanks
andnonbanks.
6See Cottarelliand Kourelis(1994) for furtherdiscussion,as well as for the
value of the structuralregressors.
FINANCIALSTRUCTURE
and the numberof bank branchesper 100,000inhabitants(NOBRA).

The expectedsign is negativefor the formervariableandpositivefor the
latter (the largerthe concentration,the lowerthe degreeof competition
and thereforethe lower the multipliers).17
Whilethis is the standardapproach,the theoryof contestablemarkets
impliesthat marketconcentrationmeasuresare not good proxiesfor the
actual degree of competition.The reason is that concentratedmarkets
can behavelike competitivemarketsif firmsare subjectto the threatof
entry of new competitors.We have thereforeincludedin the regression
a qualitativeindex of the existence of barriersto entry (ENTRY). This
index, rangingfromzero (strongestbarriersto entry)to four (no barriers
to entry), reflects the legislationon the opening of new bank branches
(bothdomesticandforeign)existingin eachcountry,andhasan expected
positive sign. Moreover,in some regressionspecifications,the variable
MARSH and NOBRA have been includedin the followingform:
MARSH* = (4 - ENTRY)*MARSH (12)
NOBRA* = (4 - ENTRY)/NOBRA (13)
Equations(12) and (13) imply that the degree of marketconcentration
becomesrelevantonly in the presenceof barriersto entry.18
The stronger
those barriers,the greaterthe impact of marketconcentrationon the
multipliers.
Extentof Money MarketDevelopmentand Opennessof the Economy
The extentto whichthe moneymarketis developedhasbeen takeninto
accountin two ways.First, a variablemeasuringthe size of the "random
component"in the moneymarketrate seriesused in the step one regres-
sors was included(RANDO). The expected sign of this variableis neg-
ative: if the money market rate series are very "noisy," the speed of
adjustmentshould be lower. This is because, in the presenceof adjust-
17Thisis becauseof the lowerdemandelasticityin less competitivemarketsand

the presumedhigherstickinessof oligopolisticprices.It hasbeen notedin Section
I that, while there are reasonsto arguethat oligopolisticpricesmay be stickier
thancompetitiveprices,the sameargumentmaynot holdfor monopolisticprices
(apart from the effect imputed to the lower demand elasticitycharacterizing
monopolies).Thus the relationbetween degree of stickinessand concentration
may not be linear. To take this into account, absolute deviationsfrom the
MARSHsamplemeanhavebeen calculated(so thatveryconcentratedandvery
fragmentedmarketswould behave similarly).However, this has not yielded
substantiallydifferentresultsfrom those reportedin Section IV.
18 Note that, in this specification,the expectedsign of NOBRA* is now nega-
tive (as NOBRA appears in the denominator),while the expected sign of
MARSH* continuesto be negative.
CARLOCOTTARELLI
ment costs, bankswill only follow interestrate changesthat are not too
erratic.RANDO has been set equal to the standarderror(expressedas
a percentage of the average value of the money market rate) of an
ARIMA model fitted on each money marketinterest rate series.19
The second aspectto be consideredis the size of the marketfor short-
term negotiablefinancialinstrumentsissued by enterprises(ENTMA)
andotheragents(OTHMA),both measuredin relationto eachcountry's
GDP.20The existence of a marketfor short-terminstrumentsissued by
enterprises(commercialpaper and bankers'acceptances)may be rele-
vant becauseit increasesthe elasticityof the demandfor bankloans. In
this case, if banks do not adjust rapidlyto changes in money market
conditions,they may be disintermediated.The existenceof a marketfor
other short-termmarketableinstruments(mainlycertificatesof deposit
(CDs) and treasurybills) may also be important.The existenceof these
instrumentsincreases the liquidity of enterpriseand household port-
folios, thus increasingthe elasticityof demandfor loans. Moreover,if
banks raise a large share of their resourcesfrom the issuanceof CDs,
whose interestratesrapidlyadjustto moneymarketconditions,they will
face large costs if they delay the adjustmentof their lendingrates.
An additionalvariable-CAPCO-has been introducedto capture
the barriersto foreigncompetition.Its expectedsignis negative;it takes
the value 1 in the presenceof constraintson capitalflows and the value
0 otherwise.21
BankingSystemOwnership
As more comprehensivemeasures of the degree of public sector
ownershipwere not readily available,the public/privatenature of the
banking system was measured by a variable (PUBLI), equal to the
19For simplicity,the same (2,1,2) ARIMA model was fitted to all series.
20
To accountfor the possibilitythatthe size of the moneymarketis not relevant
beyonda certainlevel, the abovevariableswere also introducedin the following
nonlinearform:
OTHMA*= 1/(1 + )exp(-TrOTHMA)),
that is, througha logisticfunction.This specificationimpliesthat, for very high
as well as very low levels of OTHMA, changesin the marketsize have limited
effect. The twoparameters0 andTrwereestimatedby scanning(i.e., by minimiz-
ing the residual sum of squares). The estimated ( and ir implied a close linear
relationbetween OTHMA*and OTHMA (for the actualvalues taken by the
latterin the cross-countrysample),whichsuggeststhatthe effectof OTHMAwas
approximatelylinear.
21No attempthas been made to differentiateby type of controlson capital
movements. Annual information on this variable has been derived from Alesina,
Grilli, and Milesi-Ferretti (1993).
numberof the five largestbanksthatarepublic.Thisvariableis expected

to be negativelyrelated to the impactmultipliers.
Degree of Developmentof the FinancialSystem
In order to test the hypothesis that lending rates adjust faster in
more sophisticatedfinancialenvironments,we includedvariablesmea-
suring the overall degree of development of the financial system. A
standardapproachwouldrequiretakingthe ratiobetweentotal financial
assetsandGDP. Thismeasure,unfortunately,is not readilyavailablefor
all countriesincludedin the sample.Wethereforeusedthreeproxies:per
capitaGDP (GDPPC), which usuallyexhibitsa strongcorrelationwith
the ratio between financialassets and GDP;22the ratio between broad
money and GDP (M2GDP), which is often used as a proxy for the de-
gree of financialdeepening(e.g., De GregorioandGuidotti(1992));and
the ratio between broad and narrowmoney (M20M1), which captures
the developmentof more sophisticateddeposit instruments.
AdditionalVariables
To identifythe effect of the abovefactors,it is necessaryto controlfor
the existence of other variablesinfluencingthe measuredmultipliers.
First,two dummyvariableswereintroducedto distinguishbetweenthe
typeof lendingrateusedin the step one regressions.The variablePRIME
takesthe value 1 for postedprimeratesandzero otherwise.It is expected
to havea positivesign, sinceratesappliedto the best (i.e., higherdemand
elasticity)customersarelikelyto reactfasterandbecausethe adjustment
costs for changingposted rates are lower than for changingactualrates.
The variablePOSTEtakesthe value 1 for nonprimepostedratesandzero
otherwise.Its sign is uncertainbecause the two factorsmentionedwith
referenceto the PRIME variablenow move in differentdirections.
Second, adjustmentlags of nominalvariables(nominalpricesor inter-
est rates) are likely to be shorterin environmentsin whichinflationhas
been high for a numberof yearsand, consequently,indexationis widely
used (Cecchetti(1986)).Structuralinflationwasmeasuredas the average
inflationrate duringthe 1980s (INFLA).
Third,the variableEDISC was includedto test the possibilitythat the
multiplieris lowerwhenthe discountrateis usedas a signalingdevice(the
possible "discount-rateaddiction" hypothesis noted in Section II).
EDISC, whichis definedas a dummyvariabletakingthe value 1 for the
22
Forthe 32 countriesconsideredby Wellons,Germidis,and Glavanis(1986),
the correlationcoefficientbetweenper capitaGDP and financialassetsto GDP
ratio is 0.66.
countries in which the discount rate was significantin the step one
regressions,is expectedto havea negativesignif the addictionhypothesis
is true.
Fourth, we also included an additionaldummyvariable (DUSHO)
equal to 1 for countries in which the sample period of the step one
regressionwas shorter than two years. This variablewas included be-
cause, in the presence of a lagged dependentvariable,OLS estimates,
while consistent,are biased(the so-calledHurwiczbias).As discussedin
Nickell(1981),thisbiasis likelyto resultin an overestimationof the speed
of adjustment.Therefore,we expect the sign of DUSHO to be positive.
Specification Search and Preferred Equations

ImpactMultiplierEquation
Table 4 reports the estimates of equation (5), that is, the relation
between impactmultipliersand the structuralvariables,for Model 1.23
Followingthe "from general to specific" approach,the specification
searchstartedwiththe inclusionof all exogenousvariableslistedabove.24
The estimatesof the most generalspecificationsare reportedas esti-
mates(1)-(2) in Table4, referring,respectively,to the OLSandweighted
least squares (WLS) results. While the two estimates are similar,it is
confirmedthat the use of OLS would have produced artificiallylow
coefficient standard errors (and correspondinglyhigher t-statistics).
However,even estimate(2) presentsa remarkablygood fit (the adjusted
R2 is 0.80, whichis very high for cross-sectionestimates)and low stan-
dard errors.25Of the 13 variables included in the regression, only
MARSH and GDPPC have a sign opposite to what was expected. Of
these, MARSH, which measuresthe market share of the five largest
banks,is veryclose to zero andis not significant,26
leavingpercapitaGDP
as the only significantvariablewith the "wrong"sign. As recalled,this
23
See Cottarelliand Kourelis(1994) for the resultsobtainedusing Model 2
estimates,whichwere very similar.
24However,given the limitednumberof degreesof freedom, the alternative
proxies for the degree of financialdevelopment(i.e., GDPPC, M2GDP and
M20M1) were introducedindividually.Table 4 only reports the results for
GDPPC,as M2GDPandM2OM1were neversignificant.The DUSHO variable
was also never significantand was droppedto save degreesof freedom.
25 Both the adjustedR2 and the equationstandarderrorhave been expressed
in terms of the originalresiduals,i.e., those of the estimatenot adjustedfor
heteroscedasticity(the correspondingstatisticson the equation adjusted for
heteroscedasticitylook, of course, even better).
26Forthe given sample size, a 10 percent and a 5 percentsignificancelevel
requiret-statisticsof 1.70 and 2.04, respectively.
Table 4. Estimatesof Equation (5)
(Dependent Variable: Impact Multipliers from Model
Estimation
N Constant INFLA PRIME POSTE CAPCO RANDO ENTMA OTHMA PUBLI GDPPC MA
Technique
OLS 1 0.50 0.012 0.25 -0.34 -0.18 -0.031 -0.001 0.010 -0.056 -1.29 0
(4.09) (9.81) (4.63) (-5.69) (-3.49) (-5.73) (-0.09) (3.87) (-4.61) (-2.42) (0
WLS 2 0.51 0.013 0.22 -0.37 -0.21 -0.035 -0.003 0.013 -0.039 -1.50 0
(4.35) (7.09) (3.22) (-3.93) (-3.07) (-5.45) (-0.43) (4.48) (-2.50) (-2.34) (0
WLS 3 0.30 0.012 0.22 -0.28 -0.15 -0.031 -0.002 0.011 -0.032 - 0
(3.57) (-1.90) - (0
(3.66) (6.18) (2.87) (-2.97) (-2.18) (-4.51) (-0.30)
0.25 -0.27 -0.12 -0.032 - 0.009 -0.022 - -
WLS 4 0.30 0.012
- (3.76) (-1.94) - -
(4.29) (6.85) (3.92) (-4.04) (-2.27) (-4.69)
0.012 0.25 -0.27 -0.12 -0.032 - 0.009 -0.022 - -
WLS 5 0.33
(-5.93) - (4.05) (-1.77) -
(5.17) (7.12) (5.01) (-4.04) (-2.13)
WLS 6 0.30 0.011 0.17 -0.29 -0.12 -0.025 - 0.013 -0.045 - -
- (5.48) (-5.04) - -
(4.07) (6.53) (3.50) (-3.86) (-2.01) (-4.28)
OLS 7 0.31 0.011 0.20 -0.33 -0.18 -0.027 -0.002 0.011 -0.064 -0.54 0
(1.98) (9.84) (4.58) (-5.24) (-3.36) (-4.97) (-0.27) (4.86) (-5.78) (-0.99) (1
WLS 8 0.45 0.012 0.15 -0.37 -0.21 -0.030 -0.002 0.015 -0.060 -1.33 0
(2.60) (6.52) (2.38) (-3.44) (-2.78) (-4.72) (-0.27) (5.24) (-4.68) (-1.42) (0
0.011 0.15 -0.31 -0.15 -0.026 - 0.013 -0.046 -
WLS 9 0.26
(-4.60) - (5.76) (-5.26) - -(
(3.41) (6.81) (2.97) (-4.21) (-2.45)
aIn estimates 7-9 this variable is adjusted for the existence of barriers to entry (see Se
variable acts as a proxy for the level of financial development, and thus
is not important on its own. Therefore, it was dropped in estimate 3,
without any major change in the other coefficients and t-statistics.
In estimate (3), four variables (ENTMA, MARSH, NOBRA, and
ENTRY) are not significant. Therefore, ENTMA and MARSH (the
least significant of the group) are dropped in estimate (4), which raises
the t-statistics for the remaining two variables. These, however, continue
to be insignificant. It must be noted that NOBRA (the number of bank
branches) and ENTRY (reflecting the ease of opening bank branches)
show a relatively high correlation,27so that their lower significance, when
introduced in tandem, may reflect problems of multicollinearity. Indeed,
when the two variables are introduced separately in estimates (5) and (6),
respectively, they each become significant at the 1 percent significance
level. On account of the lower standard error and higher adjusted R2,
estimate (6) will be considered the "preferred" equation.28
In estimates (7)-(9), MARSH and NOBRA are replaced by their
corresponding values adjusted for the existence of barriers to entry (see
equations (12) and (13) above), but the results do not change appreciably.
ENTMA, MARSH* and NOBRA* remain insignificant. GDPPC is also
not significant, while ENTRY is significant even in the most general
specification. This confirms estimate (6) as the preferred equation.
Interim and Long-Term Multiplier Equations
While the focus of this paper is on the impact multipliers for the reasons
discussed in Section II, it is worthwhile to examine how the estimated
equations behave when applied to interim and long-term multipliers.29
These estimates, reported in Table 5 (again, for Models 1 and 2)
together with the preferred impact equations, show a much worse fit. The
adjusted R2 drops to 0.50 and 0.23), respectively, for the three- and
six-month multipliers, and becomes negative for the long-run multi-
27Their correlationcoefficientis 0.52, meaningthat the numberof branches

is higher
28It
in countrieswith lower barriersto openingnew branches.
can be noted that the adjustedR2of estimate(6) is almostas high as that
of the overparameterized estimate(2). Thisindicatesthat the significanceof the
"wronglysigned"GDPPC may have been spurious.
29As noted in Section II, the standarderrorsof the estimatedinterim and
long-termmultipliersare not easilycomputable.Therefore,in the estimationof
the correspondingstep two equations,we adjustedfor heteroscedasticityusing
the standarderrorsof the impactmultipliers.Thisis not a problemas long as the
standarderrors of the interim multipliersare equal to those of the impact
multipliersup to a multiplicativeconstant.
30Theuse of the adjustedR2, however, underestimatesthe portion of the
dependentvariablevarianceexplainedby the equation.Even for the long-term
multiplierequation,the unadjustedR2 remainsclose to 0.30.
Table 5. Estimates of the Interim and Long-Term Multiplier
(Dependent Variable: Multipliers from Model 1)
Estimation
Technique Multiplier Constant INFLA PRIME POSTE CAPCO RANDO OTHMA P
WLS Impact 0.30 0.011 0.17 -0.29 -0.12 -0.25 0.13 -
(4.07) (6.53) (3.50) (-3.86) (-2.01) (-4.28) (5.48) (-
WLS 3-month 0.71 0.014 0.20 -0.24 -0.14 -0.036 0.021 -
(5.66) (4.77) (2.29) (-1.90) (-1.33) (-3.60) (5.33) (-
WLS 6-month 0.80 0.012 0.18 -0.19 -0.13 -0.035 0.024 -
(5.06) (3.34) (1.70) (-1.20) (-1.01) (-2.85) (4.80) (-
WLS Long-run 1.01 0.005 -0.02 -0.21 -0.23 -0.017 0.020
(6.26) (1.43) (-0.19) (-1.27) (-1.71) (-1.35) (3.95)
plier.30This is not surprising,as we noted that the variabilityof the

multipliersacrosscountriestendsto fade awayin time, so thatit becomes
more difficult (but also less relevant)to explain it. Nevertheless,it is
notable that the signs and, to some extent, the significanceof the coef-
ficients remainunchanged,particularlyup to the six-monthmultiplier
equation.
Discussion of the Econometric Results

The above resultssupportthe followingconclusions.
Effectof Inflation
The results indicatethat the speed of adjustmentof lending rates is
higherin inflationaryenvironments,a resultthatreplicatesthatobtained
for commodityprices by Cecchetti(1986). In all of the above specifica-
tions, the coefficient on INFLA is very significantand close to 0.01,
indicatingthat an increasein the structuralrate of inflationby 10 points
raisesthe impactmultiplier(andindeedthe multipliersup to six months)
by 10 basis points.
Typeof LendingRate
The resultsalsoindicatethatthe dynamicsof the adjustmentof lending
rates vary dependingon the type of lending rate. Prime posted rates
adjustfaster than actualrates (their multiplieris almost20 basis points
higher,for up to six months),while posted nonprimerates adjustmore
slowly, particularlyin the very short run (their impactmultiplieris 30
basis points lower than for actualrates, and 20 basis points lower after
three months).This impliesthat, when assessingthe effectivenessof the
transmissionmechanismof monetarypolicy, attentionmust be paid to
the type of lendingrate for whichinformationis available.Only in the
long run do all rates tend to change by the same amount.
Effect of FinancialStructure
Theeconometricresultsalsoindicatethatthe stickinessof lendingrates
is stronglyinfluencedby the structureof the financialsystem, including
its regulatoryenvironment.The effects of five structuralvariableshave
been identified (Table6).31
First, the stickinessof lendingrates has been shownto be influenced
31 Table6 suggeststhatthe effectof changesin the differentstructural

variables
canbe added.Whilethe estimatedmodelis, indeed, additive,it mustbe stressed
that additivityprobablydoes not hold for very high or very low values of the
multipliers(see Section II).
Table 6. Effectof StructuralChangeson the LendingRateMultiplier

StructuralChange Impact 3-month 6-month
Removalof barriersto entry 0.14
Privatizationof the bankingsystem 0.23 0.34 0.27
Removalof capitalcontrols 0.12
Creationof a money market
(equal to 15 percentof GDP) 0.20 0.32 0.36
50 percentreductionof "noise"on the
money marketratea 0.13 0.18 0.18
aReductionfrom10percentto 5 percentin the ratiobetweenthe standarderror
of the randomcomponentof the moneymarketrateseriesandits averagevalue.
by the existence of constraints on competition among banks, and in

particular, by the existence of barriers to entry (measured here by con-
straints in setting up new bank branches).32 Based on the estimated
regression coefficients, a shift from a regime of ad hoc authorization in
the opening of branches to one of complete deregulation is estimated to
increase the impact multiplier by 14-19 basis points.33The actual degree
of concentration (measured by the market share of the five largest banks)
seems to be less relevant. This is consistent with the view, stressed by the
contestable market school, that very concentrated markets behave like
competitive markets as long as they are subject to entry threat.34
Second, lending rates appear to be stickier in banking systems
dominated by state banks, which may reflect the relative inefficiency of
public banks or the existence of political constraints on interest rates
changes. Privatizing a publicly owned banking system would substantially
increase the flexibility of lending rates. The impact multiplier would be
raised by over 20 basis points, and the effect would be even higher for
the three- and six-month multipliers.
Third, capital controls reduce competitive pressures on the banking
32
As mentioned,similarresultshavebeen obtainedby usinga measureof the
actualdiffusionof bank branches.
33 At higherorderlags, the effectsare less clearlyidentified(the corresponding
t-statisticsare low) althoughthe size of the estimatedcoefficientremainshighup
to the six-monthmultiplier.
34 The existenceof barriersto interstatebranching,and hence to competition,
wouldbe one factorexplainingthe relativelyhighdegreeof stickinessof lending
ratesin the United States (Table1), despitethe low degreeof marketconcentra-
tion. The sameconclusionholdsfor ItalyandJapan,whichin the sampleperiod
maintainedstrong barriersto the opening of new branches.In contrast, the
Canadianbankingsystem,whichis very concentratedbut characterizedby rela-
tively low entry barriers,exhibits a faster adjustment.For a more detailed
discussionon the relationbetweenentrybarriersand competitionin the United
States and Canada,see Shaffer(1993).
system (arising from foreign financial markets) and result in higher

lendingrate stickiness.The quantitativeeffect of removingcapitalcon-
trols, while significantfor the impactmultiplier,is relativelycontained
(12 basis points), and is statisticallyinsignificantafterward.However,it
must be recalledthat the capitalcontrolvariablehas been measuredin
a very impreciseway, which may explain the relativelyhigh standard
errorof the correspondingcoefficient.35
Fourth,the developmentof a marketfor short-terminstruments(par-
ticularlyCDs and treasurybills) also enhancesthe flexibilityof lending
rates. For a marketas largeas, say, 15 percentof GDP, the effect would
be between20 and30 basispointson all multipliersup to six months.We
were unable to identifyany effect of marketsfor short-termnegotiable
instrumentsissued by enterprises. One possible interpretationis that
these instruments(particularlycommercialpaper) are issued mainlyby
very large enterprises,while in manycountries,the bulk of commercial
bank loans is granted to medium-sizedand small enterprisesand to
households.
Fifth, quiteintuitively,lendingratesdo not followmoneymarketrates
that move very erratically.If the ratiobetweenthe standarderrorof the
randomcomponentof the moneymarketrateandthe averageof the same
ratedeclinesby 5 percentagepoints,the multipliersincreasesubstantially
(10-20 basis points depending on the lag and the model). Thus, the
growthof the money marketcan speed up the responseof the banking
system by reducingthe volatilityof the money marketrate (under the
assumptionthat interestrate volatilityis, ceterisparibus,lowerin larger
markets). In general, the transmissionmechanismwill benefit from
avoidingexcessive fluctuationsof money marketrates.
Role of the DiscountRate
One featureof the regressionspresentedin Tables4-5 is the statistical
significance,and the negativesign, of the coefficientreflectingthe dis-
count rate policy of the centralbank. The estimatedcoefficientimplies
that the use by the central bank of the discount rate as a monetary
policy signalreducesthe responseof lendingrates to changesin money
market rates (when the discount rate is not moved) by 15-30 basis
points (depending on the lag and model specification).The fact that
this result has been obtained after controllingfor a large number of
structuralvariables affecting the stickiness of lending rates supports
the "discount-rateaddiction" hypothesis put forward at the end of
Section III.
35Indeed, while the t-statisticsof CAPCOfall afterthe impactmultiplier,the

estimatedcoefficientremainshigh.
T-able7. UnitedKingdom and Canada:Estimatesof the Lending
(In percent)
Sampleperiod N Constant i m m- Ad d
United Kingdom
72:10-78:03 1 0.72 0.63 0.77 -0.41 -0.01 ---
(2.10) (6.68) (7.46) (-2.80) (-0.09) --
78:04-81:02 2 0.10 0.99 -0.01 0.95 ---
(0.73) (39.52) (0.11) (0.38) (32.73)

78:04-81:02 3 0.98 0.03 --- 0.97
(5.26) (0.15) - - (5.40)
81:03-93:03a 4 0.87 0.38 0.86 -0.23 - ---
(5.16) (6.04) (18.82) (-3.53) - --
Canada
73:01-80:02 5 0.14 0.93 0.01 0.05 0.95 ---
(1.28) (34.98) (0.09) (0.53) (8.58) -
73:01-80:02 6 0.12 0.95 --- 0.93

(1.20) (28.79) - - - (16.73)
80:03-91:10 7 0.13 0.76 0.01 0.26 0.79 ---
(1.45) (17.51) (0.14) (4.51) (13.11) -
aA dummyvariablein January1985 was also included (see Appendix III in Cottarellia

616 CARLOCOTTARELLI
and ANGELIKIKOURELIS
It couldbe arguedthat, basedon the estimatedcoefficienton EDISC,

the stickiness attributedto discount-rateaddictionsis relatively con-
tained, andthat it is a reasonablepriceto payfor an effectiveinstrument
such as an administrativelycontrolleddiscountrate. However, the dis-
countrateis an effectiveinstrumentonlyinsofaras it canbe flexiblyused.
But, as arguedabove, administeredratesmaybe relativelysticky.More-
over, the estimatedeffect of the discount-rateaddictionreportedabove
reflects the average response of the banking systems included in the
sample, and it may thereforeunderestimatethe effect in specificcoun-
tries. Furtherevidence on this point can be derived by reviewingthe
experienceof two countriesin whichthe discountratewasused, but only
for some periods, as an administeredsignalingdevice.
Table7 focuseson the relationbetweenthe lendingrate, moneymarket
rates, and the discountrate in the United Kingdomand in Canada.In
the United Kingdom,betweenOctober13, 1972,andApril 11, 1978,the
discountrate (that is, the MinimumLendingRate of the Bank of Eng-
land, or MLR)was set at 0.5 percentabovethe averagetreasurybill rate
at the most recent tender (Temperton(1991), p. 162), and thus did not
have any independentsignalingeffect. As indicatedby the firstestimate
of the table, the lendingrate in this period was primarilyinfluencedby
the moneymarketrate, witha relativelyshortadjustmentlag (the impact
multiplieris 0.77). The MLR was administeredbetweenApril 11, 1978,
and August 20, 1981. Clearly, in this period, the relevanceof money
marketratesdropped(equation(2)), andthe MLRbecamethe reference
rate for banks. Indeed, the lending rate adjusts to the MLR almost
simultaneously(equation(3)). Whilethis maybe believedto be an ideal
conditionfor a centralbank, Temperton(1991) notes that:
Disenchantmentwiththisregimesoon set in. Changesin the officialinterest
rateonce againtook a highpoliticalprofileandthis led to problemswiththe
conductof monetarypolicy.... On 20 August, 1981,it was statedthat MLR
would no longer be announcedcontinuously:greaterreliance was to be
placedon marketforcesin the determinationof interestrates,... (p. 163).
Equation(4) shows that, after the suspensionof the MLR in August
1981, money marketrates once again became the main determinantof
lendingrates, with very short adjustmentlags.36
36In the late 1980s, a new administered rate (the so-called Band One
Stop Rate,
whichis the minimumrate at whichthe Bank of Englandis willingto discount
bills of less than 14 days maturity)graduallyemergedas a signalingdevice of
monetary policy changes. This rate has been shown to affect money market and
bank interestrates quite rapidly(Dale (1993)). The differencescomparedwith
the MLR are that the changes in the Band One Stop Rate, while closely moni-
tored by financial markets, do not receive the same attention by the media and
have a lower political impact, and therefore can be used more flexibly.
Canada'sexperiencewas similar.Until March1980,the discountrate

was set administrativelyand playedthe role of signalingchangesin the
stanceof monetarypolicy (Freedmanand Dingle (1986),p. 28). Before
that date, money marketrates did not appearto influencelendingrates
(equation (5)). Indeed, the level of the lending rate appeared to be
related uniquelyto the level of the discountrate (equation(6)). In the
followingperiod,the discountratewasindexedto the levelof the treasury
bill rate, thus losingits role as a policysignal.As illustratedby equation
(7), duringthe 1980s, lendingrates were still statisticallyrelated to the
discountrate, nowto be interpretedas a proxyof the mostrecenttreasury
bill auctionrate (see Cottarelliand Kourelis(1994), Appendix III).
These resultsconfirmthe quantitativerelevanceof the discount-rate
addiction hypothesis. When the discount rate is used as a signaling
device, banks become less reactiveto money marketchangesthat are
unaccompaniedby discountrate changes.37
V. Conclusionsand Policy Implications

The stickinessof lendingrateswithrespectto changesin moneymarket
rates has often been seen as a seriousimpedimentto the smooth trans-
mission of monetarypolicy impulses. Yet, no systematicattempt had
previouslybeen made to measurethe differentdegree of stickinessof
lendingratesacrosscountriesor to explainthe observeddifferences.This
paperhas attemptedsucha measurementand, by doingso, has provided
a yardstickagainstwhich the degree of lending rate stickinessin indi-
vidualcountriescan be assessed. It has shownthat the degree of sticki-
ness is quite different across countries, particularlyin the very short
run. The impact multiplier(defined as the change in the lending rate
observedduringthe month in whichthe money marketrate changes)is
close to unity in some countries (i.e., the adjustmentis completed in
almost one month) but as low as zero in others. Significantdifferences
can still be observedafter three and six months, while, in the long run,
the adjustmenttends to be close to unity for most countries.
The paper has also documentedthe existence of a strong relation
between the degree of interest rate stickinessand the structureof the
financialsystem. Five structuralfeatureshave been singledout as being
particularlyrelevantin increasinglendingrateflexibility:the existenceof
37 Admittedly,the aboveresultsmayoverstatethe loss of significanceof money

marketratesin the presenceof an administereddiscountrate. Mostlikely,banks
wouldstopusingthe discountrateas a referencerateif the latterweremaintained
excessivelyout of line with respectto money marketrates. In this respect,it is
interestingto note that between 1978 and 1981, the Bank of Englandkept the
MLR relativelyclose to money marketrates (Spencer(pp. 55-57)).
618 CARLO COTIARELLI and ANGELIKI KOURELIS
a marketfor negotiable short-terminstruments(particularlyCDs and

treasurybills);the containmentof "unnecessary"or randomfluctuations
in moneymarketrates;the absenceof constraintson internationalcapital
movements;the absence of constraintson bank competition(particu-
larly, barriersto entry); and privateownershipof the bankingsystem.
Marketconcentrationandthe existenceof marketsforinstrumentsissued
by enterprises(for example,commercialpaper)did not appearto affect
loan rate stickiness. These results were obtained after controllingfor
structuralinflation(whichtends to speed up the adjustmentof lending
rates) and for the type of lendingrates used (posted primerates adjust
fasterthan actualrates, whichin turnreactfasterthannonprimeposted
rates).
These resultsadd a new dimensionto the relationbetweenregulation
policiesandmonetarypolicy.The analysisof thisrelationhas, in the past,
focusedon the aspectof "soundness,"that is, on the fact that the finan-
cial system must be resilientenough to sustainstrongmonetarypolicy
measures"untilthey beginto bite"(Revell(1980),Gardener(1978)).We
focusedprimarilyon the relationbetweencompetitionandefficiency,on
the one hand, and monetarypolicy, on the other. Based on our results,
the transmissionmechanismof monetarypolicy can be enhanced by
policiesaimedat enrichingthe financialstructureof new markets(partic-
ularlyfor short-termmarketableinstruments),andby removingbarriers
to competition (such as barriersto entry and constraintson capital
movements).Privatizationpolicies also appearto affect the responsive-
ness of lendingratesto monetarypolicystimuli,possiblybecauseprivate
banks are more efficient, or because they are less subject to political
constraints.
Policies aimed at reducingmarketconcentrationdo not appearto be
useful, possiblybecausecompetitionis best guaranteedby the threatof
entry, both on local and nationalmarkets,ratherthan by increasingthe
numberof nationalcompetitors.Therefore,policiesfavoringbankmerg-
ers, such as those implementedby some Europeancountriesin the last
few years, may not be inconsistentwith competition.
It has been shownthat the presenceof a high level of noise in money
market rates weakens their role as conveyersof monetarypolicy im-
pulses, possiblyby makingit more difficultfor banksto identifydurable
changesin interestrates.Theremaythereforebe a caseforpoliciesaimed
at smoothingmoney marketrate fluctuations.38
The above results also have implicationsfor the shift from direct to
indirectmonetarycontrols.Direct creditceilingswere commonin many
38These policies may include structural regulatory changes, such as reserve

averaging and lagged reserve requirements.
FINANCIALSTRUCTURE
industrialcountriesduringthe 1960sand 1970s,and are still widelyused

in developingcountries.As noted in a BIS reportof the early 1970s:
... quantitativecreditceilings... are seen to havethe advantageof helping
to limitthe growthof creditandthe moneysupplymorequicklyandprecisely
than would be possible by the use of conventionalmonetaryinstruments
actingthroughbank liquidityand interestrates (BIS (1971), p. 1).
One of the reasonsfor the system'slimitedresponsivenessto changes
in indirectmonetaryinstrumentsis the stickinessof banklendingrates.
However,as arguedabove,thisstickinessshouldnot be takenfor granted
as it is influencedby factorsthat can be modifiedby structuralreforms.
Thus, before ruling out the possibilityof shiftingto indirectcontrols,
considerationshould be given to structuralreformsaimed at enhancing
the transmissionmechanismof indirectmonetaryinstruments.
Finally,the paperalso has implicationsfor the use of the discountrate
as a monetarypolicyinstrument.By signalingfundamentalchangesin the
monetarypolicystance, administrativechangesin the discountratestim-
ulate the responseof lendingratesto moneymarketchanges.Therefore,
in countriesin which lendingrates are sticky due to the weaknessesof
the financialstructure,there is a strongcase for using an administered
discountrate as part of the centralbank policy arsenal,until the effect
of structuralfinancialreformsgraduallybeginsto bite. At the sametime,
evidence has been presentedsupportingthe so-calleddiscount-ratead-
dictionhypothesis,namelythat the repeateduse of the discountrate as
a policy signalweakensthe responseof banksto moneymarketchanges
that are not accompaniedby discount rate changes. This may be a
problemfor monetarypolicybecauseadministereddiscountratesmaybe
more easily subjectto politicalpressuresof variousforms, and present
some degree of stickiness. Thus, in countriesin which the structural
barriersto lendingrate flexibilityhave been removed,there is a case for
de-emphasizingthe discountrateas policysignal,for example,by linking
it to money marketrates.
APPENDIX
Relation Between the y Coefficientsat Different Lags

As discussed in Section II, the y coefficients, expressingthe relationship
between structuralvariablesand "multipliers"at differenttime lags, are not
independentacrosslags. To explorethisrelationship,let us considerthe simplest
partialadjustmentmodel for the lendingrate:39
i= lii_l + P2m (A.1)
39All of the followingequationsshouldbe consideredas referringto a single
country;for simplicitythe subscripti used in Section II has been dropped.
620 CARLOCOTTARELLI
andANGELIKIKOURELIS
wherei is the lendingrate, andm is the moneymarketrate.The impactmultiplier

(ho)and the interimmultipliersup to lag 2 (hl, h2) 40can be expressedin terms
of p coefficientsas:
ho= P2 (A.2)
hl= P2(1+ P1) (A.3)
h2= P2(1 + P1 + P12). (A.4)
Consistentlywith equation(4) in Section II, the multipliersare expressedas a
functionof the structuralvariables(two in this example),denoted as zl and z2:
ho = yolzl + 'o2Z2 + Eo (A.5)
hi = yllZl + y12Z2+ E1 (A.6)
h2 = Y21Zl + Y22Z2 + E2, (A.7)
whereEo,El, andE2arethe errorterms.By combiningequations(A.2)-(A.4) with
(A.5)-(A.7) the relationbetweenthe p and the y coefficientscan be writtenas
follows:
P2 = YOlZ1+ Y02Z2 + Eo, (A.8)
P2(1 + Pi) = yllZl + '12Z2 + El, and (A.9)
P2(1 + P1+ P12)= Y21Zl+ '22Z2+ 62. (A.10)
Using vector notation, equations(A.8)-(A.10) can be rewrittenas:
2 = Z'yo + eo, (A.11)
P2(1+ Pi) = Z'yi + e1, and (A.12)
P2(1+ P + P12)= Z'y2 + e2, (A.13)
where yo= [YO1,Y02]',y/ = [y11,Y12]',Y2= [Y21,Y22]',
and Z' = [z1,z2]. From equa-
tions (A. 11) and (A. 12),the followingrelationbetween yoandl/ can be derived:
(Z'yo + Eo)(l + pi) = Z'yi + 61; (A.14)
and from (A.12), (A.13), and (A.14):
(Z'yo + EO)- (Z'yi + e1) + [(Z'yi + el)2/(Z'yo + eo)] = Z'Y2 + E2. (A.15)
Equation(A. 15) showsthatthe relationbetween yo, yl, and Y2does not involve
anyfurtherinformationon the ps. However,the elementsof Y2cannotbe derived
from /yoandyl becausethe constraintset by (A.15) is on linearcombinationsof
the elements of the y vectors and not on the elementsof the vectors.
REFERENCES
Alesina, Alberto, VittorioGrilli,and GianMariaMilesi-Ferretti,"ThePolitical

Economy of CapitalControls,"CEPR WorkingPaperNo. 793 (London:
Centrefor EconomicPolicy Research,June 1993).
40Westop at lag 2 for simplicity'ssake. The algebrabecomes increasingly

complicatedat longer lags.
Bankfor InternationalSettlements,QuantitativeCreditRestrictions(Basle:BIS,
1971).
Bernanke,Ben S., "Creditin the Macroeconomy,"FederalReserveBankof New
YorkQuarterlyReview,Vol. 18 (Spring1993),pp. 50-70.
, andAlan S. Blinder,"Credit,Money,and AggregateDemand,"Amer-
ican EconomicReview,Papersand Proceedings,Vol. 78 (May 1988), pp.
435-45.
Bernanke,Ben S., and MarkGertler,"AgencyCosts, Net Worth,and Business
Fluctuations,"American Economic Review, Vol. 79 (March 1989) pp.
14-31.
Cecchetti, Stephen G., "The Frequencyof Price Adjustment:A Study of the
NewsstandPricesof Magazines,"Journalof Econometrics,Vol. 31 (April
1986),pp. 255-74.
Cottarelli,Carlo, and Angeliki Kourelis, "FinancialStructure,Bank Lending
Rates,and the Transmission Mechanismof MonetaryPolicy,"IMFWorking
Paper94/39 (Washington:InternationalMonetaryFund, March1994).
Dale, Spencer,"TheEffectof Changesin OfficialU.K. Rateson MarketInterest
RatesSince1987,"Papersin Money,MacroeconomicsandFinance,Supple-
mentto The ManchesterSchool, Vol. 61 (June 1993),pp. 76-94.
De Gregorio,Jose, and Pablo E. Guidotti, "FinancialDevelopmentand Eco-
nomic Growth,"IMF WorkingPaper 92/101 (Washington:International
MonetaryFund, December 1992).
Faig-Aumalle,Miquel, Implicationsof BankingMarketStructurefor Monetary
Policy, IMF WorkingPaper 87/25 (Washington:InternationalMonetary
Fund, April 1987).
Freedman,C., and J.F. Dingle, "MonetaryPolicyImplementationin Canada:
TraditionalStructureand Recent Developments,"in Changesin Money-
MarketInstrumentsand Procedures:Objectivesand Implications,by Bank
for InternationalSettlements,MonetaryandEconomicDepartment(Basle:
BIS, March1986),pp. 23-40.
Gardener,E. P. M., "ThePhilosophyof Bank CapitalAdequacy,"in Competi-
tionandRegulationof Banks, ed. by JackRevell,BangorOccasionalPapers
in Economics,No. 14 (Cardiff:Universityof WalesPress,1978),pp. 24-33.
Gilbert,R. Alton, "BankMarketStructureand Competition:A Survey,"Jour-
nal of Money, Creditand Banking, Vol. 16 (November1984, Part 2), pp.
617-45.
Graybill,FranklinA., An Introductionto LinearStatisticalModels(New York:
McGrawHill, 1961).
Hannan,TimothyH., and Allen N. Berger,PriceRigidityand MarketStructure:
Theoryand Evidencefrom the BankingIndustry,Financeand Economics
DiscussionSeries, No. 59, (Washington:Boardof Governorsof the Federal
Reserve System, March1989).
, "TheRigidityof Prices:Evidencefromthe BankingIndustry,"American
EconomicReview,Vol. 81 (September1991),pp. 938-45.
Harvey,A.C., TheEconometricAnalysisof TimeSeries(Cambridge,Massachu-
setts: MIT Press, 1990).
Heggestad, Arnold A., "A Survey of Studies on Banking Competitionand

Performance,"in Issuesin FinancialRegulation,ed. by FranklinR. Edwards
(New York:McGraw-Hill,1979),pp. 449-90.
Kareken,John H., "Bank Regulationand the Effectivenessof Open Market
Operations,"BrookingsPaperson EconomicActivity:2 (1984),The Brook-
ings Institution,pp. 405-55.
Klemperer,Paul, "Marketswith ConsumerSwitchingCosts,"QuarterlyJournal
of Economics,Vol. 102 (May 1987), pp. 375-94.
Klein, MichaelA., "A Theoryof the BankingFirm,"Journalof Money, Credit
and Banking,Vol. 3 (May 1971),pp. 205-18.
Maddala,G. S., Econometrics(New York:McGraw-Hill,1979).
Modigliani,Franco,andLucasPapademos,"TheStructureof FinancialMarkets
and the MonetaryMechanism,"in ControllingMonetaryAggregatesIII,
Proceedingsof a ConferenceSponsoredby the FederalReserveBank of
Boston (Boston: Federal Reserve Bank of Boston, October 1980), pp.
111-55.
Monti, Mario, "A TheoreticalModel of Bank Behaviorand its Implicationsfor
MonetaryPolicy,"L'Industria,Vol. 0 (April/June1971),pp. 165-91.
, "Deposit, Credit, and InterestRate DeterminationUnder Alternative
Bank Objective Functions,"in MathematicalMethodsin Investmentand
Finance, ed. by G.E. Szergo and K. Shell (Amsterdam:North-Holland,
1972),pp. 430-54.
Nickell, Stephen, "Biasesin DynamicModels with Fixed Effects," Economet-
rica, Vol. 49 (November1981),pp. 1417-26.
Pelzman,Sam, "The BankingStructureand the Transmissionof MonetaryPol-
icy," Journalof Finance,Vol. 14 (June 1969),pp. 387-411.
Revell, Jack, "The ComplementaryNatureof Competititionand Regulationin
the FinancialSector,"in CompetititionandRegulationin FinancialMarkets,
ed. by A. Verheirstraeten(London:Macmillan,1980).
Rodrigues, Anthony, "GovernmentSecurities Investments of Commercial
Banks," FederalReserveBank of New York QuarterlyReview, Vol. 18
(Summer1993),pp. 39-53.
Saxonhouse,GaryR., "EstimatedParametersas DependentVariables,"Amer-
ican EconomicReview,Vol. 66 (March1976),pp. 178-83.
, "Regressionsfrom SamplesHavingDifferentCharacteristics,"Review
of Economicsand Statistics,Vol. 59 (May 1977) pp. 234-37.
Shaffer, Sherrill,"A Test of Competitionin CanadianBanking,"Journalof
Money, Creditand Banking,Vol. 25 (February1993),pp. 49-61.
Short, Brock K., "The Relation between CommercialBank Profit Rates and
BankingConcentrationin Canada,WesternEurope,andJapan,"Journalof
Bankingand Finance,Vol. 3 (September1979),pp. 209-19.
Spencer,Peter D., FinancialInnovation,Efficiency,and Disequilibrium:Prob-
lemsof MonetaryManagementin the UnitedKingdom,1971-1981 (Oxford:
ClarendonPress, 1986).
Temperton,Paul, U.K. MonetaryPolicy: The Challengefor the1990s(London:
Macmillan,1991).
Tseng, Wanda,and Robert Corker,FinancialLiberalization,Money Demand,

and MonetaryPolicy in Asian Countries,IMF OccasionalPaper No. 84
(Washington:InternationalMonetaryFund, July 1991).
VanHoose,DavidD., "BankMarketStructureandMonetaryControl,"Journal
of Money, Creditand Banking, Vol. 17 (August 1985),pp. 298-311.
Wellons,Philip,DimitriGermidis,and BiancaGlavanis,Banksand Specialized
FinancialIntermediaries
in Development(Paris:DevelopmentCentreof the
Organizationfor EconomicCooperationand Development,1986).

Financial Structures, Bank Lending Rates and Transmission Mechanism

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Financial Structures, Bank Lending Rates and Transmission Mechanism

Uploaded by

Copyright:

Available Formats

Financial Structure, Bank Lending Rates, and the Transmission Mechanism of Monetary

Financial Structure, Bank Lending

T of monetary policy hinges on a set of crucial

markets. These structuralparameters(such as the elasticities of the

'See, for example, Modiglianiand Papademos (1980), Vanhoose (1985),

cial structure,we follow a simpleapproach.First,we measurethe speed

I. The Stickinessof Bank LendingRates

Definition of Lending Rate Stickiness

The term "interestrate stickiness"has taken on two related, but dis-

4 Thus, this paper is clearly related to the "bank structure-performance" liter-

in the long run (long-runstickiness).In this paperwe will refer mainly

The Relevance of the Financial Structure

The term "financialstructure"will be used fairlybroadlyto include

costs, however,the lendingrate will changeonly if those costs are lower

II. The Empirical Model

whereii,t,mitanddi,,are, respectively,the lendingrate,the moneymarket

ii = PO(1 - P1) + [(P2 + *' + Pn+2)/(1 - Pl)]m, (2)

whichis consistentwith the monopolisticcompetitionmodel relatingthe

hi,e = 4([i), (3)

hi,e = ZiYe+ vi,e, (4)

ho = Zyo + vo; (5)

Discussion of the Model

Definitionof the Multipliers

Suppose that the 13coefficientshad been modeled directlyas a linear

The DynamicSpecificationof the Model

the impactmultiplierho = 13o(whose varianceis estimateddirectlyfrom

III. Step One: Analysis of Stickiness

Havinglaid out the schemeof the empiricalmodel, we can proceedto

9High frequencydata are needed because time aggregationmay bias the

central bank10and no direct controlson the amountof credit were in

Table 1 shows the estimatedmultipliersof changesin money market

10In orderto allowsomeinitialadjustmentof the

cient (that is, the ratio between standarddeviationand mean) of each

countries,the averageimpactmultiplierof a changein the discountrate

IV. Step Two:Determinants

Steptwo-the estimationof the relationbetweenmultipliersandstruc-

and the numberof bank branchesper 100,000inhabitants(NOBRA).

17Thisis becauseof the lowerdemandelasticityin less competitivemarketsand

numberof the five largestbanksthatarepublic.Thisvariableis expected

Specification Search and Preferred Equations

27Their correlationcoefficientis 0.52, meaningthat the numberof branches

plier.30This is not surprising,as we noted that the variabilityof the

Discussion of the Econometric Results

31 Table6 suggeststhatthe effectof changesin the differentstructural

Table 6. Effectof StructuralChangeson the LendingRateMultiplier

by the existence of constraints on competition among banks, and in

system (arising from foreign financial markets) and result in higher

35Indeed, while the t-statisticsof CAPCOfall afterthe impactmultiplier,the

(2.10) (6.68) (7.46) (-2.80) (-0.09) --

78:04-81:02 2 0.10 0.99 -0.01 0.95 ---

(0.73) (39.52) (0.11) (0.38) (32.73)

(5.16) (6.04) (18.82) (-3.53) - --

(1.28) (34.98) (0.09) (0.53) (8.58) -

73:01-80:02 6 0.12 0.95 --- 0.93

(1.45) (17.51) (0.14) (4.51) (13.11) -

aA dummyvariablein January1985 was also included (see Appendix III in Cottarellia

It couldbe arguedthat, basedon the estimatedcoefficienton EDISC,

Canada'sexperiencewas similar.Until March1980,the discountrate

V. Conclusionsand Policy Implications

37 Admittedly,the aboveresultsmayoverstatethe loss of significanceof money

a marketfor negotiable short-terminstruments(particularlyCDs and

38These policies may include structural regulatory changes, such as reserve

industrialcountriesduringthe 1960sand 1970s,and are still widelyused