Embed Size (px)
Citation preview
This article was downloaded by: ["University at Buffalo Libraries"]On: 07 October 2014, At: 11:38Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK
Applied Financial EconomicsPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/rafe20
Modelling bank lending in the euro area: a nonlinearapproachLeonardo Gambacorta a & Carlotta Rossi ba Monetary Economics Department , Bank for International Settlements , Basel, Switzerlandb Structural Studies Department , Bank of Italy , Rome, ItalyPublished online: 01 Jul 2010.
To cite this article: Leonardo Gambacorta & Carlotta Rossi (2010) Modelling bank lending in the euro area: a nonlinearapproach, Applied Financial Economics, 20:14, 1099-1112, DOI: 10.1080/09603101003781430
To link to this article: http://dx.doi.org/10.1080/09603101003781430
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
Applied Financial Economics, 2010, 20, 1099–1112
Modelling bank lending in the euro
area: a nonlinear approach
Leonardo Gambacortaa,* and Carlotta Rossib
aMonetary Economics Department, Bank for International
Settlements, Basel, SwitzerlandbStructural Studies Department, Bank of Italy, Rome, Italy
This article investigates the possible nonlinearities in the response of bank
lending to monetary policy shocks in the euro area. The credit market is
modelled over the period 1985 to 2005 by means of an Asymmetric Vector
Error Correction Model (AVECM) involving four endogenous variables
(loans to the private sector, real Gross Domestic Product (GDP), lending
rate and consumer price index) and one exogenous variable (money market
rate). The main features of the model are the existence of two cointegrating
equations representing the long-run credit demand and supply and the
possibility for loading and lagged term coefficients to assume different
values depending on the monetary policy regime (easing or tightening).
The main result of this article is that the effect on credit, GDP and prices of
a monetary policy tightening is larger than the effect of a monetary policy
easing. This finding supports the existence of an asymmetric broad credit
channel in the euro area.
I. Introduction
There is a widespread consensus among economists
and policy-makers on the leading indicator properties
of credit regarding medium-term prospects for output
and inflation.1 The role of credit in the conduct of
monetary policy is in fact emphasized by the practice
of several central banks of including credit in their
information set. The European Central Bank (ECB)
explicitly considers credit in both pillars of its
monetary policy strategy as ‘given the particular
importance of bank loans for the financing of euro
area firms, developments in such loans may have
important implications for euro area-wide economic
activity’ (ECB, 2004). The Federal Reserve, too,
assigns a special role to credit as ‘policymakers
continue to use monetary and credit data as a
source of information about the state of the economy’
(Bernanke, 2006). Both central banks use the bank
lending survey to obtain detailed information on
developments in financial conditions.From a theoretical perspective, the role of the
credit market in the transmission mechanism of
monetary policy is emphasized by authors adhering
to the credit view. Although the credit channel was
studied by Hawtrey, Hahn, Keynes and other authors
of the ‘Swedish School’ in the 1930s, the debate about
its relevance has been re-opened by the work of
Bernanke and Blinder (1988). According to these two
authors, the credit market is characterized by imper-
fect substitutability between bank loans and
privately-issued debt. Indeed, where some borrowers
*Corresponding author. E-mail: [email protected], amongst others, Stock and Watson (1999), Friedman and Kuttner (1993), Kashyap et al. (1993) for the US andNicoletti-Altimari (2001) and Calza et al. (2006) for the euro area. Stiglitz and Greenwald (2003) demonstrate that lendingconditions help us to understand two recent historical episodes: the East Asian financial crisis and the 1991 US recession andsubsequent recovery and boom.
Applied Financial Economics ISSN 0960–3107 print/ISSN 1466–4305 online � 2010 Taylor & Francis 1099http://www.informaworld.com
DOI: 10.1080/09603101003781430
Dow
nloa
ded
by [
"Uni
vers
ity a
t Buf
falo
Lib
rari
es"]
at 1
1:38
07
Oct
ober
201
4
do not have access to the capital markets (not onlyhouseholds, but even small firms), their spending andinvestment decisions must rely exclusively on theavailability of bank credit and self-financing: in sucha case, every change in the composition of bank assetswould influence both the level and the distribution ofprivate expenditure for consumption and investment.
Within the credit view, two transmission channelshave received most attention: the ‘Bank LendingChannel’ (BLC) and the ‘Balance Sheet Channel’(BSC). The BLC claims that a monetary tighteningaffects bank loans because the drop in reservabledeposits cannot be completely offset by issuing otherforms of funding (i.e. uninsured CDs; for an oppositeview see Romer and Romer, 1990) or by liquidatingsome assets. Imperfect substitutability between bankloans and other forms of funding for firms (equityand bonds) produces a negative effect on investmentand output growth.
Loan supply shifts could also originate from a‘BSC’, working through the relative prices of theguarantees provided to banks (Mishkin, 1995; Olinerand Rodebusch, 1996; Kashyap and Stein, 1997): amonetary squeeze increases debt service costs, whichcan prompt sales of real assets, reducing their valueand causing a loss of creditworthiness and a reduc-tion of lending. In this situation, there is a greaterincentive for banks to finance less risky projects andto start a ‘flight to quality’ (Lang and Nakamura,1995). The ‘financial accelerator hypothesis’(Bernanke and Gertler, 1989; Bernanke et al., 1996)reinforces this mechanism, claiming that a monetarypolicy tightening, by reducing the net worth ofborrowers, increases the premium on external financerequired by lenders, thus reducing borrowers’ abilityto invest. These mechanisms are amplified if loandemand also responds to the changes in monetarypolicy and output asymmetrically, due to a differen-tial effect on investment decisions and self-financing(Friedman and Kuttner, 1993). In this case, a ‘broadcredit channel’ is identified because both supply anddemand schedules move asymmetrically in responseto a monetary change.
A natural implication of the credit view is that thepass-through of monetary policy shocks to outputand inflation may be asymmetric. From a theoreticalperspective, there are various ways of rationalizingthe nonlinear behaviour of credit markets in the caseof monetary policy changes. The most obvious one isthe outright assumption that the loan supply curvehas some form of rigidity. The first example is themodel of Stiglitz and Weiss (1981) in which anadverse selection problem leads to a backwardbending supply of credit and a consequent creditrationing on the upside. Blinder (1987) builds a model
in which monetary policy shocks have differenteffects when the economy is in a credit rationingregime with respect to other more accommodativeregimes. McCallum (1991) finds evidence that thecredit-rationing mechanism exists and that it helps toexplain the fact that in the US, during the 1980s, theoutput effect of money shocks was about twice aslarge in monetary policy tightening than in easing.
Asymmetric behaviour may also be detected evenif no rationing a la Stiglitz and Weiss is detected andloan supply always matches loan demand. Forexample, considering the BLC mechanism, we maythink that, due to capital regulation, it is impossiblefor a bank to further increase the supply of new loansin the case of a monetary easing. This is because it isnecessary to maintain a certain proportion betweenbank capital and lending. This rigidity does not applyin the case of a monetary tightening, and a reductionin economic activity is usually associated with adrop in loan demand. In this case, the reduction ofsupplied credit frees up capital requirements andmakes it possible to cover any losses. De Long andSummers (1988) interpret the asymmetric mechanismin a different way, claiming that ‘banks can eitherremain healthy or they can fail. If banks fail there arenegative macroeconomic ramifications but there areno corresponding possibilities on the positive side’.
Nonlinear dynamics are also detected in the ‘BSC’and the ‘financial accelerator theory’. For example,Bernanke et al. (1996) stress that the changes in creditmarket conditions amplify and propagate the effectsof initial real or monetary shocks, thus, explaining thesmall shocks/large cycles puzzle. The intrinsicallynonlinear nature of the financial accelerator theory isalso strengthened by moral hazard and adverseselection problems that amplify the effect of amonetary policy shock via credit more in a recessionthan during a boom.
In this article, we examine empirically whethercredit in the euro area reacts nonlinearly to monetarypolicy shocks. This topic is particularly critical for theeuro area economy in virtue of the importance ofbank financing. At the end of 2005, the total amountof bank lending to euro area households and firmsamounted to 114% of Gross Domestic Product(GDP), compared with 58% in the US. Even ifempirical findings support the existence of a ‘creditchannel’ in the transmission mechanism of monetarypolicy in the euro area (see Angeloni et al., 2003, fora survey) all the studies have been performedusing linear regressions or linear autoregressions(Vector Autoregression, VAR) that may not detectnonlinearities in the credit market as depicted in thecredit view literature. Therefore, the main novelty ofthis article lies in the test for asymmetric adjustment
1100 L. Gambacorta and C. Rossi
Dow
nloa
ded
by [
"Uni
vers
ity a
t Buf
falo
Lib
rari
es"]
at 1
1:38
07
Oct
ober
201
4
of bank loans, real GDP and prices in the event ofopposite monetary policy impulses (tightening oreasing). The econometric framework used is theAsymmetric Vector Error Correction Model(AVECM) as in Lim (2001). This is based on areformulation of the multivariate error correctionmodel proposed by Johansen (1988, 1995), whichallows for an asymmetric behaviour both in the longand in the short run. In particular, the model capturesthe interplay of long-run relationships (a loandemand and a loan supply), embedded in the co-integration vectors, with their short-run adjustmentscaptured by the part in first difference.
The article is organized as follows. Section IIreviews the empirical literature on the modelling ofcredit to the private sector in the euro area.Section III gives a descriptive analysis of the dataand identifies possible breaks in the estimationperiod. After an examination of the characteristicsof the VAR model in Section IV, Section V discussesthe long-run relationship between bank lending,GDP, consumer prices and interest rates usingJohansen’s methodology. Section VI presents theAVECM used to test for the presence of asymmetricbehaviour depending on whether policy rates areincreasing or decreasing. Model specification tests arereported in Section VII, while Section VIII containsthe results of a simulation using the estimatedAVECM. The last section summarizes the mainconclusions.
II. Modelling Credit to the Private Sector
Since the beginning of the century, modelling creditto the private sector has been the subject of a fastgrowing strand of empirical literature.2
There exist a disagreement in this literature con-cerning the role of the supply side of the creditmarket. The majority of work focuses on the demandside of the credit market under the assumption thatthe supply effects are limited (in line with the money
view). According to the credit view, however, mon-etary policy affects the real economy by shifting thesupply schedule of credit, and the overall effects ofthe BLC and BSC are considered to work in additionto the traditional interest rate channel. Several studieshave attempted to disentangle the different channelsby focusing on cross-sectional differences betweenbanks.3 This microeconometric literature, however,considers linear mechanisms where monetary tight-ening and easing have similar impact in magnitude onsupplied lending. Some models depart from thisassumption and analyse the role of credit as anonlinear propagator of shocks (Balke, 2000, forthe US; Atanasova, 2003, for the UK; Calza andSousa, 2006, for the euro area). These authorsgenerally employ a VAR model including output,inflation, credit and a policy rate. The focus is on theimpact of monetary policy shocks on output andinflation, allowing VAR coefficients to vary depend-ing on the credit market regime (high lending rate orlow lending rate regimes).
In this article, we examine the nonlinear nature ofthe credit market from a different perspective.According to the credit view, in fact, credit is likelyto react directly in a nonlinear way to different typesof policy shocks (easing or tightening). We thereforeanalyse, by means of an AVECM, the reaction ofcredit, output and inflation to positive or negativepolicy shocks, assuming the monetary policy rate asthe transition variable.
In other words, we allow credit not only to capturethe asymmetric propagation of monetary policy, butalso to react directly to policy shocks in a nonlinearway.4
III. Data Description
This study is based on quarterly data for the euroarea over the period 1985:01 to 2005:04. The inter-action between the credit market and economicactivity is analysed by means of the following
2See Kakes (2000), Huelsewig (2003), Casolaro and Gambacorta (2005), Calza et al. (2006) and Casolaro et al. (2006).For a review of the literature on credit market models see, amongst others, the working paperversion of this studyhttp://www.bancaditalia.it/pubblicazioni/econo/temidi/td07/td650/en_tema_650.pdf.3See, amongst others, Altunbas et al. (2002), Ehrmann et al. (2003) and Gambacorta (2005).4A further perspective in the analysis of the asymmetric behaviour of credit market is provided by Granger and Lee (1989)that investigate asymmetric adjustments in dynamic economic relationships. According to their methodology, the speed ofadjustment in a Vector Error Correction Model (VECM) model is allowed to depend on whether the deviation of theendogenous variables from the equilibrium level indicated by the co-integrating vector is positive or negative. We applied thetest suggested by Granger and Lee to our framework to investigate if our results could be affected also by this kind ofasymmetric adjustment. However, the test is always rejected suggesting that credit does not react in a nonlinear way topositive/negative values of the co-integrating vectors (see Cook et al., 1999, for the statistical properties of this test). Theresults are available upon request.
Modelling bank lending in the euro area 1101
Dow
nloa
ded
by [
"Uni
vers
ity a
t Buf
falo
Lib
rari
es"]
at 1
1:38
07
Oct
ober
201
4
variables: credit to the private sector (c), real GDP
(y), the consumer price index (p), the average lending
rate (l) and the monetary policy interest rate (i).5
Figure 1 shows a high correlation between the
GDP and credit suggesting the possibility that they
are co-integrated. Better economic conditions usually
increase the number of projects becoming profitable
in terms of expected net present value and hence
increase the demand for credit (Kashyap et al., 1993).
This calls for the presence of a long-run relationship
between the credit and GDP. However, as stressed by
Friedman and Kuttner (1993), the impact of business
cycles on consumption and investment may also be
countercyclical due to a self-financing effect that
reduces the proportion of bank debt.Figure 2 shows that the spike in nominal credit
growth in the late 1980s may be due, at least in part, to
the high inflation rate. Nominal lending is probably
positively correlated with the price index, and the
hypothesis of homogeneity between the two variables
will be formally tested in the econometric section.
Figure 3 depicts the composite lending rate and thepolicy rate. The behaviour of the composite lendingrate and of the monetary policy indicator clearlyshows the two series to be co-integrated. However,there is a break in the difference between the twoseries l – i (the mark-up) that typically captures bothcredit risk and structural characteristics of the lendingmarket (Fig. 4). In other words, the mark-up tends toincrease if, other things being equal, borrowersbecome more risky or if conditions on the creditmarket alter. Examples may be changes in creditsupply towards forms of credit that are less guaran-teed or more short term. The increase in the mark-upoccurred from the fourth quarter of 1992 to the firstquarter of 1995. During this period – characterized bythe European Monetary System (EMS) crisis, a dropin the real GDP and stock market capitalization – thequality of bank lending decreased sharply. A largenumber of firms defaulted and the delinquency ratefor consumer credit peaked. All these elements pointto a change in the risk component in the mark-uptowards a new equilibrium level.6
2
4
6
8
10
12
14
–6
–4
–2
0
2
4
6Credit to the private sector (left-hand scale)
Real GDP (right-hand scale)
1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005
Fig. 1. Credit and the business cycle (quarterly data, annual percentage changes)Source: ECB, Eurostat and national statistics.
5Nominal credit is measured by the outstanding amounts of loans to the private sector in EU-12 countries. The real GDPseries for EU-12 countries is provided by Eurostat, while consumer price index is obtained by the Organisation for EconomicCooperation and Development (OECD) statistical database. All these series are reported in logs and are seasonally adjusted.The composite lending rate is calculated as a weighted average of rates on different types of loans (see Calza et al., 2006, forfurther details). The policy rate is the Euro Overnight Index Average (EONIA) interest rate.6This break is not due to a shift in the maturity composition of bank credit (we checked for this by analysing only short-termcredit). The increase in the mark-up was more pronounced in France and Germany, where the drop in the quality of creditand in bank profitability in 1992 to 1995 was particularly large, calling for an increase in risk premia. It is worth mentioningthat, on the contrary, the mark-down, the difference between the monetary policy indicator and the interest rate on deposits,declined substantially in all the major euro area countries over the sample period. This indicates that the conditions applied todepositors have improved, probably due to greater competition in the deposit market. Further details on the data are reportedin the working paper version of the study (http://www.bancaditalia.it/pubblicazioni/econo/temidi/td07/td650/en_tema_650.pdf).
1102 L. Gambacorta and C. Rossi
Dow
nloa
ded
by [
"Uni
vers
ity a
t Buf
falo
Lib
rari
es"]
at 1
1:38
07
Oct
ober
201
4
IV. The Vector Error Correction Model
The five variables are modelled using a VAR system;all the variables are found to be I(1) without drift. TheI(1) nature of the variables calls for the existence of oneor more co-integrating relationships. Conventionaltests point to the presence of two co-integratingvectors (Table 1) and show that, according to eco-nomic theory, the monetary policy rate (i) can beconsidered weakly exogenous (Table 2).
The model may then by represented by the follow-
ing way:
Dxt ¼ �ð�, xt�1, it�1Þ þXp�1k¼1
�kDxt�k þXp�1k¼0
�kDit�k
þ �dum90t þ "t t ¼ 1, . . . ,
T� ¼ ��0 ð1Þ
where p¼ 3; xt¼ [c, y, l, p]; � is an n� r matrix
of loading coefficients and � is an n� r matrix of
0
2
4
6
8
1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005
Fig. 2. Consumer price index (quarterly data, annual percentage changes)
Source: OECD.
0
2
4
6
8
10
12
14
1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005
Composite lending rate
Money market interest rate
Fig. 3. Interest rates (quarterly data, percentage values)
Source: Calza et al. (2006) and ECB (2006).
Modelling bank lending in the euro area 1103
Dow
nloa
ded
by [
"Uni
vers
ity a
t Buf
falo
Lib
rari
es"]
at 1
1:38
07
Oct
ober
201
4
co-integrating vectors (n¼ 4 and r¼ 2); dum90 rep-resents a point dummy included to reach a normalresiduals distribution; and the constant is included inthe co-integration space.
According to the economic theory, the twoco-integrating vectors could be interpreted as a loandemand and loan supply equations. Loan demandshould be a positive function of real GDP and prices
1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 20050
1
2
3
4
5
Mark-upDummy modelling
Fig. 4. Structural break in the mark-up (quarterly data)Source: Calza et al. (2006) and ECB (2006) and authors’ calculations.
Table 1. Co-integration analysis
H0: r¼ 0 H0: r� 1 H0: r� 2 H0: r� 3 H0: r� 4
Trace statistic 122.55** 64.01** 32.28 9.44 3.665% critical values 75.1 52.6 33.8 19.1 7.3
Constrained co-integrating vectors (SEs in brackets)
c ¼ 1:310ð0:14Þ
y� 4:327lð0:56Þ
þpþ 5:396ð1:133Þ
l ¼ iþ 0:016ð0:01Þ
c1 þ 0:005ð0:00Þ
c2 þ 0:026ð0:00Þ
c3
Loading coefficients (SEs in brackets)
� ¼
�0:05ð0:02Þ
0:03ð0:14Þ
�0:06ð0:05Þ
�0:18ð0:22Þ
�0:01ð0:01Þ
�0:22ð0:06Þ
0:04ð0:04Þ
0:10ð0:09Þ
2666664
3777775
Notes: Test for the co-integration rank of the models. Johansen �-trace tests take into account the adjustment for degrees offreedom proposed by Reimers (1992) for small samples. In order to make the test consistent with the presence of a pointdummy in the model, asymptotic critical values, reported below, are calculated with the approach suggested by Johansen andNielsen (1993). The asymptotic estimates of the critical values involve 100 000 simulations with 800 steps in the approximationto Brownian motion.** Denotes rejection at the 1% significance level.
1104 L. Gambacorta and C. Rossi
Dow
nloa
ded
by [
"Uni
vers
ity a
t Buf
falo
Lib
rari
es"]
at 1
1:38
07
Oct
ober
201
4
and a negative function of the composite lendingrate. In other words, we suppose the existence of alog-linear long-run relationship of the typec¼�1,1 yþ �1,2 lþ �1,3 p, in which the hypothesis ofhomogeneity between loans and prices may be testedfor �1,3¼ 1.
As for loan supply, economic theory on oligopo-listic (and perfect) competition suggests that in thelong run, the lending rate should be related to themonetary policy rate that represents the cost ofbanks’ refinancing. For example, Freixas andRochet (1997) show that in a model of imperfectcompetition among N banks, each one sets itslending rate as the sum of the exogenous money mar-ket rate and a constant mark-up. The supply scheduleis of the form l¼ iþ� where � ¼ �c þ l0ðc�Þc�=N isconstant. The mark-up � is influenced by risk andconstant marginal cost of intermediation on lend-ing �c and by the elasticity of the loan demandfunction evaluated at the optimum (c* is the amountof credit at the equilibrium). It is worth noting thatin the case of perfect competition (N!1), thelast part of � goes to zero and the mark-upis only influenced by marginal costs and the riskcomponent.
As discussed in the previous section, the mark-up isnot constant in our sample period. Figure 4 showsthat � is stationary in 1985–1992 and in 1995–2005,but it exhibits a trend in the period 1993 to 1994.In order to make the mark-up interpretable, we there-fore allow the constant in the co-integratingspace to assume different values over the twosub-periods 1985:01–1992:03 and 1995:01–2005:04.Moreover, we replace the constant with a time trendfor the period that goes from the fourth quarter of1992 to the last quarter to 1994 in order to approx-imate the catching up process towards the newequilibrium.
The normalized co-integrating relationships arepresented in the second part of Table 1. The set of
overidentified restrictions, including the test for pricehomogeneity in the loan demand equation (�1,3¼ 1),is accepted with a p-value of 7%.
As for the estimated coefficients, the long-runelasticity between lending and GDP, �1,1, is equal to1.3, which is consistent with Calza et al. (2006) whoobtain a long-run elasticity of 1.5 using a similar setup. Income elasticity above one is likely to reflect theomission of some variables from the models such aswealth or house purchases that are not captured byGDP transactions. The semi-elasticity of bank loanswith respect to the composite lending rate �2 isnegative (�4.3) in line with the findings of Calza et al.(2006). As for the second co-integrating vector, themark-up � is equal to 1.6% over the period1985:01–1992:03 and 2.6% during the period1995:01–2005:04.
V. The Asymmetric Vector ErrorCorrection Model
The model analysed so far is symmetric. However, asdiscussed in Section I, credit is likely to react in anonlinear way to positive and negative policyshocks via the BLC or the BSC. This impliesasymmetric adjustment of loans in both magnitudeand speed, and therefore the multivariate frame-work described by (1) should be extended to allowfor asymmetric behaviour in the loading coeffi-cients (�) and in the lagged responses of variables indelta (�).
Following Saikkonen (1992), Lim (2001),Calza and Zaghini (2006) and Gambacorta andIannotti (2007), who use a similar framework,we assume that the intercepts and the elasticities ofthe long-run relationships do not vary across differ-ent monetary policy regimes. This means that inthe long run, the equilibrium in the credit marketis unique.
Preliminarily, we test whether it is worthwhileto move from the symmetric model to the more gen-eral asymmetric model. According to Terasvirtaet al. (1994), we carry out a LR test, where thesymmetric (asymmetric) model is the restricted(unrestricted) one. This test is particularly usefulbecause, if no significant gain is detected using themore general model, it is possible to stop furtherinvestigation. The test confirms7 the need for anasymmetric approach to the problem.
The VECM system (1) with two co-integratingvectors can be reformulated as follows:
Table 2. Weak exogeneity
H0: weakexogeneity of LR-test p-value
c 17.75 0.00y 9.20 0.01l 7.46 0.02p 22.96 0.00i 4.43 0.11
Note: Weak exogeneity is accepted when the p-value islarger than 5%. LR: Likelihood Ratio.
7The LR statistic is equal to 105.28. It has a chi-squared distribution with 52 degrees of freedom. The p-value is 0.00.
Modelling bank lending in the euro area 1105
Dow
nloa
ded
by [
"Uni
vers
ity a
t Buf
falo
Lib
rari
es"]
at 1
1:38
07
Oct
ober
201
4
Dct¼ð�c,1þ��c,1dtÞ ½ct�1�pt�1þ�1,1yt�1þ�1,2lt�1þ�1�
þ ð�c,2þ��c,2dtÞ½lt�1� it�1þ�2,1c1þ�2,2c2þ�2,3c3�
þXp�1k¼1
ð�c,kþ ��c,kdtÞDct�kþ
Xp�1i¼1
ð’c,kþ’�c,kdtÞDyt�k
þXp�1k¼1
ð c,kþ �c,kdtÞDlt�kþ
Xp�1k¼1
ðc,kþ �c,kdtÞDpt�k
þXp�1k¼0
ð#c,kþ#�c,kdtÞDit�kþ�cdum90þ"ct ð2Þ
Dyt¼ ð�y,1þ��y,1dtÞ½ct�1�pt�1þ�1,1yt�1þ�1,2lt�1þ�1�
þ ð�y,2þ��y,2dtÞ½lt�1� it�1þ�2,1c1þ�2,2c2þ�2,3c3�
þXp�1k¼1
ð�y,kþ ��y,kdtÞDct�kþ
Xp�1i¼1
ð’y,kþ’�y,kdtÞDyt�k
þXp�1k¼1
ð y,kþ �y,kdtÞDlt�kþ
Xp�1k¼1
ðy,kþ �y,kdtÞDpt�k
þXp�1k¼0
ð#y,kþ#�y,kdtÞDit�kþ�ydum90þ"yt ð3Þ
Dlt¼ ð�l,1þ��l,1dtÞ½ct�1�pt�1þ�1,1yt�1þ�1,2lt�1þ�1�
þ ð�l,2þ��l,2dtÞ½lt�1� it�1þ�2,1c1þ�2,2c2þ�2,3c3�
þXp�1k¼1
ð�l,kþ ��l,kdtÞDct�kþ
Xp�1i¼1
ð’l,kþ’�l,kdtÞDyt�k
þXp�1k¼1
ð l,kþ �l,kdtÞDlt�kþ
Xp�1k¼1
ðl,kþ �l,kdtÞDpt�k
þXp�1k¼0
ð#l,kþ#�l,kdtÞDit�kþ�ldum90þ"lt ð4Þ
Dpt¼ ð�p,1þ��p,1dtÞ½ct�1�pt�1þ�1,1yt�1þ�1,2lt�1þ�1�
þ ð�p,2þ��p,2dtÞ½lt�1� it�1þ�2,1c1þ�2,2c2þ�2,3c3�
þXp�1k¼1
ð�p,kþ ��p,kdtÞDct�kþ
Xp�1i¼1
ð’p,kþ’�p,kdtÞDyt�k
þXp�1k¼1
ð p,kþ �p,kdtÞDlt�kþ
Xp�1k¼1
ðp,kþ �p,kdtÞDpt�k
þXp�1k¼0
ð#p,kþ#�p,kdtÞDit�kþ�pdum90þ"pt ð5Þ
In this AVECM, the two co-integrating vectors are
normalized on c and l, respectively. The constant
term in the first long-run relationship (loan demand)
is �1, and in the second co-integrating vector (loan
supply), the mark-up �2 is divided into three parts
(�2,1, �2,2 and �2,3) in order to capture the structural
changes discussed in the previous sections (the
dummy variables c1 and c3 take the value 1 in the
periods 1985:01–1992:03 and 1995:01–2005:04 andzero elsewhere; c2 is a time trend only for the period1992:04–1994:04 and is fixed at zero elsewhere).
The parameters that refer to asymmetric behaviourare those with the superscript ‘*’. These are interactedwith the dummy variable d, which captures thedifferential effects of increases and decreases inthe monetary policy indicator. There are two possiblestances of monetary policy: monetary easing (a neg-ative change in the rate) and monetary tightening(a positive change in the rate). Therefore, d is definedaccording to the following scheme:
d ¼1 if DiM 5 0
0 if DiM 4 0
�
In a few cases, no quarterly changes are detected inthe monetary indicator (DiM¼ 0). In these quarters,a monetary easing (tightening) is considered,d¼ 1(d¼ 0), if the 3-month Euribor rate decreases(increases), leading to easier (more difficult) access tointerbank liquidity.
VI. Testing Asymmetry and the Reduced-Form Model
According to Lim (2001), the tests for asymmetryhave been carried out considering the null hypothesisof zero restrictions on the dummy variables inEquations 2–5. Table 3 summarizes the results.
The test for asymmetry in the loading coefficients(see Panel A of Table 3) supports the hypothesis of adifferent adjustment to disequilibrium gaps only inthe case of the lending supply curve relationship inthe price equation (��p,2 6¼ 0). The economic interpre-tation of this asymmetry is that, if an exogenousshock on supplied lending makes the interest rate onloans different from the long-run equilibrium, thereadjustment process of this co-integrating vector viaprices takes place only in the case of a monetarytightening. On the contrary, in the case of a ‘loose’monetary policy, the loading coefficient is closeto zero and there is no readjustment at all(�p,2 þ �
�p,2dt ffi 0). This behaviour calls for the exis-
tence of a ‘BSC’ that is transmitted from asset pricesto consumer prices due to ‘moral hazard’ and‘adverse selection’ problems. In other words,credit–supply disequilibria contribute to lower pricesduring a monetary policy tightening, while there areno (upside) effects on prices from the credit–supplyside during an easing regime.
It is worth mentioning that the model can befurther simplified, because the first co-integrating
1106 L. Gambacorta and C. Rossi
Dow
nloa
ded
by [
"Uni
vers
ity a
t Buf
falo
Lib
rari
es"]
at 1
1:38
07
Oct
ober
201
4
vector does not enter the lending equation and thesecond co-integrating vector does not directly affectthe credit equation. The hypothesis �1,1¼�
�1,1¼ 0
implies that, given banking costs and risk premia,banks’ price adjustments are driven by monetaryconditions. The hypothesis �c,2¼�
�c,2¼ 0 implies that
the quantity of credit in the long run is demanddriven.8
The test for asymmetry in the lagged terms (seePanel B of Table 3) gives information about thedynamic path of adjustment in the short run. Themain result is that asymmetries in the lagged termscannot be set aside as they help to explain a large partof the overall asymmetry in the model. This role ofasymmetries in the lagged terms is in line with thefindings of Calza and Zaghini (2006), who use asimilar framework to estimate the demand for M1 in
the euro area. The results show that credit reacts
asymmetrically to short-term changes in real GDP. In
the case of a monetary ‘easing’ leading to an increase
in real GDP, there is, other things being equal, a
reduction in bank lending due to an increase in self-
financing. This is consistent with Melitz and Pardue
(1973) and Friedman and Kuttner (1993) who stress
that the temporary increases in income are typically
associated with a self-financing effect that reduces the
proportion of bank debt. There is also an asymmetry
in the autoregressive part of the equation for GDP,
which tends to smooth the effects of cumulative
changes in real GDP in the case of a monetary easing.No significant asymmetries are detected in the
equation for the lending rate, calling for the absence
of particular frictions on banks’ price-setting behav-
iour. The lack of asymmetric effects is expected in the
long run because of the mark-up equation that
establishes a unique long-run relationship between
the monetary market indicator and the interest rate
on lending (Freixas and Rochet, 1997; Lim, 2001).
Nevertheless, in the short run, if a pure BLC via loan
supply is at work, we should identify a greater
reaction not only of quantities but also of banks’
prices in the case of a monetary tightening. However,
coefficients #�p,k turned out to be positive, but low and
not statistically significant.There are three possible explanations for the
absence of such asymmetric effects on the interest
rate on loans. First, as detected in some econometric
works for Italy (Gambacorta, 2005; Gambacorta and
Iannotti, 2007), asymmetric monetary policy effects
on lending rates vanish after some months, so that the
quarterly frequency of our data is not sufficient to
capture this mechanism. Second, banks may tighten
nonprice elements of borrowing such as real and
personal guarantees, the loan-to-value ratio or other
additional charges and commissions. Third, an asym-
metric movement in lending demand may wash out
the effect via loan supply. For example, if the
monetary tightening causes a severe reduction in
viable investment projects or a sharp reduction in
firms’ self-financing (as discussed regarding the pres-
ence of coefficient ’�2 in Equation 4), these effects may
produce a downward shift of demanded lending that
counterbalances the effect on prices caused by the
reduction in supplied lending, exacerbating the over-
all effects on quantity.Table 4 presents the results for the reduced four-
equation system, including the significant asymmetric
short-term effects.
Table 3. Tests for asymmetry
1985:01–2005:04
LR-statistic p-value
Panel A. Testing asymmetry in the loading coefficients��c, 1 ¼ �
�c, 2 ¼ 0 0.65 0.72
��y, 1 ¼ ��y, 2 ¼ 0 2.77 0.25
��1, 1 ¼ ��1, 2 ¼ 0 1.55 0.46
��p, 1 ¼ ��p, 2 ¼ 0 14.10 0.00***
Panel B. Testing asymmetry in the short-term terms��c, 1 ¼ �
�c, 2 ¼ 0 4.15 0.13
’�c, 1 ¼ ’�c, 2 ¼ 0 6.50 0.04**
�c, 1 ¼ �c, 2 ¼ 0 0.29 0.86
�c, 1 ¼ �c, 2 ¼ 0 0.84 0.66
�c, 0 ¼ �c, 1 ¼
�c, 2 ¼ 0 2.95 0.40
��y, 1 ¼ ��y, 2 ¼ 0 3.17 0.20
’�y, 1 ¼ ’�y, 2 ¼ 0 7.66 0.02**
�y, 1 ¼ �y, 2 ¼ 0 3.85 0.15
�y, 1 ¼ �y, 2 ¼ 0 0.09 0.96
�y, 0 ¼ �y, 1 ¼
�y, 2 ¼ 0 1.98 0.58
��1, 1 ¼ ��1, 2 ¼ 0 4.33 0.12
’�1, 1 ¼ ’�1, 2 ¼ 0 3.67 0.16
�1, 1 ¼ �1, 2 ¼ 0 0.21 0.90
�1, 1 ¼ �1, 2 ¼ 0 0.67 0.72
�1, 0 ¼ �1, 1 ¼
�1, 2 ¼ 0 5.23 0.16
��p, 1 ¼ ��p, 2 ¼ 0 1.61 0.45
’�p, 1 ¼ ’�p, 2 ¼ 0 9.11 0.01***
�p, 1 ¼ �p, 2 ¼ 0 13.88 0.00***
�p, 1 ¼ �p, 2 ¼ 0 6.22 0.04**
�p, 0 ¼ �p, 1 ¼
�p, 2 ¼ 0 4.14 0.25
Notes: LR: Likelihood Ratio.** and *** denote significance at the 5 and 1% levels of thenonlinear coefficient. The asymmetries are rejected whenthe p-value is larger than 0.05.
8The likelihood ratio test for �c,2 ¼ �l,1 ¼ ��c,1 ¼ �
�1,1¼ 0 is given by �2(4) ¼ 4.07 with a p-value of 40%.
Modelling bank lending in the euro area 1107
Dow
nloa
ded
by [
"Uni
vers
ity a
t Buf
falo
Lib
rari
es"]
at 1
1:38
07
Oct
ober
201
4
VII. A Simulation: Adjustment to Positiveand Negative Shocks
In order to evaluate the effect of an exogenous
monetary policy shock, a simulation exercise is
performed to generate time paths for the endogenousvariables. Fig. 5 shows the adjustment paths of
lending, real GDP, composite lending rate and
consumer prices to positive and negative changes inthe money market rate. In particular, the policy
experiment consists of increasing (decreasing) the
money market rate by 25 bp, starting from a base,where the four-equation system is in equilibrium. To
make simulations more graphically comparable, the
effects for easy monetary policy have been multipliedby �1. The main results are the following.
In line with the credit view, the final effect of
monetary policy on credit, real GDP and, particu-
larly, prices9 is larger in tight regimes than duringeasing. In particular, the effect on prices is consistent
with the existence of a broad credit channel that
works via a reduction of (asset) prices: a monetarysqueeze increases debt service costs which can prompt
sales of real assets, reducing their value and causing a
loss of creditworthiness and a drop in lending. Thisresult is in line with the ‘pushing on the string’ view ofmonetary policy (De Long and Summers, 1993;Karras, 1996), which explains the asymmetric effectsof money on output and prices by arguing thatnegative money-supply shocks affect aggregate
demand more than do positive shocks. It is worthmentioning that our result adds to this strand ofliterature by stressing the fact that the asymmetricmovement of aggregate demand depends on a differ-ent adjustment of the credit market to negative andpositive monetary policy shocks.
This result can also be explained easily using asimple graphical representation of Bernanke and
Blinder’s model. Using the CC line to indicate thecontemporaneous equilibrium on the goods and creditmarkets and LM for the equilibrium in the moneymarket, the economic system can be represented usinga simple CC–LM, AD–AS scheme (Fig. 6). Forexample, an unexpected restrictive monetary policynot only moves the LM curve to LMT, but also
produces a shift of the CC towards CCT, since moneycontraction works through the broad credit channel aswell, reducing the loans available to the economy.
Table 4. Coefficients of the model in reduced form
Equations
Coefficients c y l p
�1 �0.046 (0.01)*** �0.051 (0.01)*** 0.021 (0.00)***�2 �0.120 (0.13) �0.196 (0.06)*** 0.262 (0.09)***��2 �0.263 (0.10)**�1 4.342 (0.68)*** 4.342 (0.68)*** 4.342 (0.68)*** 4.342 (0.68)***�2,1 �0.015 (0.00)*** �0.015 (0.00)*** �0.015 (0.00)*** �0.015 (0.00)***�2,2 �0.005 (0.00)*** �0.005 (0.00)*** �0.005 (0.00)*** �0.005 (0.00)***�2,3 �0.027 (0.00)*** �0.027 (0.00)*** �0.027 (0.00)*** �0.027 (0.00)***�1,1 �1.179 (0.09)*** �1.179 (0.09)*** �1.179 (0.09)*** �1.179 (0.09)***�1,2 5.461 (0.52)*** 5.461 (0.52)*** 5.461 (0.52)*** 5.461 (0.52)***�1 0.417 (0.11)*** 0.190 (0.03)***�2 0.035 (0.12)***’1 0.088 (0.08)’2 �0.043 (0.11) �0.395 (0.14)***’2 �0.266 (0.12)** 0.446 (0.15)*** 1 0.177 (0.31) 0.123 (0.10) 0.049 (0.12) 2 �0.077 (0.09)1 0.224 (0.16) 0.660 (0.17)***2 0.647 (0.15)*** 0.640 (0.18)*** 0.367 (0.13)***0 0.394 (0.12)*** 0.474 (0.04)***1 �0.036 (0.14) 0.033 (0.06) �0.056 (0.08)
Notes: The reduction from the general model represented by Equations 4–7 to this reduced-form model hasbeen carried out according to Lim (2001). Asymmetric coefficients, as from Table 7, have been removedfrom the reduced model when statistically not significant (10% level). SEs in brackets. Coefficients and SEsfor the dummies are not reported.** and *** denote significance at the 5 and 1% levels, respectively.
9The larger nonlinear effect found for prices with respect to real GDP is in line with the results in Calza and Sousa (2006) whouse a different approach to analyse credit regimes.
1108 L. Gambacorta and C. Rossi
Dow
nloa
ded
by [
"Uni
vers
ity a
t Buf
falo
Lib
rari
es"]
at 1
1:38
07
Oct
ober
201
4
In the case of a monetary easing that moves theLM curve to LME, if the CC shift is not verypronounced, the EE equilibrium is found to havea lower impact than in the case of a tightening, bothon income ð yE � y�
�� ��5 y� � yT�� ��Þ and on prices
ð pE � p��� ��5 p� � pT
�� ��Þ.
VIII. Robustness Checks
The robustness of the results has been checked inseveral ways. The first test was to analyse the effectsof a temporary 25 bp change in money market rateson bank lending, real GDP, the composite lendingrate and consumer prices. Part B of Table 5 reportsthe effects of this policy experiment after 2, 5 and 30years, while part A presents, for the sake of compar-ison, the results obtained in the case of a permanent
change (discussed in Section VIII). As expected,‘the pushing on a string’ result still holds in theadjustment towards the new equilibrium: the effect ofmonetary policy on credit, real GDP and prices is stillgreater under tight monetary policy regimes thaneasy ones.
Second, we have analysed the implications ofendogenous adjustment of the monetary policy indi-cator (Table 5, Panel C). The assumption, made inFig. 5, of an exogenous and permanent change in themoney market rate is obviously not realistic. We havetherefore relaxed this hypothesis by allowing themoney market rate to react to output and inflationaccording to a simple Taylor rule.10 When a reactionfunction for the policy-maker is accounted for, theasymmetry in the response of credit (both real andnominal) and prices to policy shocks is smaller thanin the case without an interest rate reaction function,but is still present. The rationale for this is that themovements in the monetary policy rate only partially
–9
–8
–7
–6
–5
–4
–3
–2
–1
0
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 64 68
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 64 68
Credit TCredit ECredit
–2.5
–2
–1.5
–1
–0.5
0
0.5
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 64 68
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 64 68
GDP TGDP EGDP
0
0.05
0.1
0.15
0.2
0.25
0.3
clr Tclr Eclr
–4
–3.5
–3
–2.5
–2
–1.5
–1
–0.5
0
0.5Price TPrice EPrice
Fig. 5. Adjustment paths to positive and negative changes in the monetary policy indicator (percentage values)Note: The experiment consists of evaluating the effect after 70 quarters of a 25 bp permanent change in the monetary policyindicator. The symbol T represents a monetary tightening and E, a monetary easing. To make simulations more comparable,the effects for easing monetary policy are multiplied by �1.
10The Taylor rule takes the form it ¼ r þ �*þ�(�t � �*) þ �(yt � y*), where r denotes the real interest rate (by assumptionconstant), �* ¼ Dp* the central bank’s inflation objective and yt � y* is the output gap, i.e. the percentage deviation of actualoutput from potential output. According to Taylor (1993) and Alesina et al. (2001), we set the coefficients r ¼ 2, �* ¼ 1.5,� ¼ 1.5 and � ¼ 0.5. The annual increase of potential output is set equal to 2.25%. The values of � and � chosen to calibratethe Taylor rule are consistent with many of the empirical rules for Germany and Europe estimated using pre-Economic andMonetary Union (EMU) data.
Modelling bank lending in the euro area 1109
Dow
nloa
ded
by [
"Uni
vers
ity a
t Buf
falo
Lib
rari
es"]
at 1
1:38
07
Oct
ober
201
4
offset the asymmetric behaviour in the transmis-sion mechanism. In other words, even when thepolicy-maker is free to react to macroeconomicfluctuations, the response of output and prices to apolicy shock is still larger in a tightening than inan easing regime.
IX. Conclusions
The evidence presented in this article supports thehypothesis that tight monetary policy has a largereffect on bank lending, real output and prices thaneasy monetary policy.
From the point of view of theory, this resultsuggests the existence of a ‘broad credit channel’ inthe euro area: a rise in short-term interest rates makes
loan supply shifts more pronounced because themoral hazard and adverse selection problems increasethe need for a bank to reduce credit risk andrebalance loan portfolios via a ‘flight to quality’effect; moreover, the loan demand may also respondasymmetrically to changes in monetary policy due toa differential effect on investment decisions and self-financing. These mechanisms are in line with the‘pushing on the string’ view of monetary policy,
Table 5. Response of the endogenous variable to the changes in the monetary policy indicator (percentage values and basis
points)
25 bp increase 25 bp decrease
Series 2 years 5 years 30 years 2 years 5 years 30 years
Panel A: Permanent changec �0.98 �3.80 �5.87 0.59 1.76 3.94y �0.28 �0.97 �1.57 0.07 0.53 1.53l 0.23 0.25 0.25 �0.23 �0.25 �0.25p �0.18 �0.89 �2.83 �0.04 0.09 0.77
Panel B: Temporary changec �0.69 �0.97 �0.01 0.36 0.38 0.00y �0.20 �0.24 0.00 0.11 0.16 0.00l 0.03 0.00 0.00 �0.03 0.00 0.00p �0.10 �0.30 0.03 0.01 0.06 0.00
Panel C: Taylor rulec �0.58 �0.44 �0.04 0.39 0.24 0.02y �0.11 �0.10 �0.01 0.18 0.11 0.01l �0.05 0.01 0.00 0.06 0.00 0.00p �0.07 �0.18 �0.03 0.01 0.06 0.01i �0.05 0.02 0.00 0.13 0.00 0.00
Notes: In this table, we carry out three simulation exercises. Panel A summarizes the responses of the endogenousvariables to a 25 bp permanent change in the exogenous money market interest rate. Panel B shows how the samevariables react to a 25 bp temporary (1 year) change in the money market interest rate. Lastly, Panel C shows the reactionof the endogenous variables after a temporary 25 bp change in the exogenous money market rate when a Taylor ruleis accounted for.
E*
LME
ET
ET
ADT
y
y
EEE*
CCT
pE
p*
LMTi
CCE
AD*
p
EE
yEy*yT
ADE
AS
LM
pT
CC*
yEy*yT
Fig. 6. ‘Pushing on a string effect’ inside Bernanke and
Blinder’s model
1110 L. Gambacorta and C. Rossi
Dow
nloa
ded
by [
"Uni
vers
ity a
t Buf
falo
Lib
rari
es"]
at 1
1:38
07
Oct
ober
201
4
which claims the existence of asymmetric aggregatedemand movements depending on the sign of themonetary impulse.
With regard to policy, the ‘pushing on a string’result reinforces the case for a conservative centralbank that only cares about inflation (Rogoff, 1989).The lexicographical order in the ECB statute infavour of inflation is preferable because any attemptof the monetary authority to increase output has alimited effect. It is worth mentioning that this result isderived in a more general framework than theKeynesian liquidity trap.
Further research could be directed towards twoadditional issues. First, the analytical framework ofthe model could be used to analyse possiblenonlinearities in the response of bank lending tomonetary policy shocks in single European countries,checking whether the degree of asymmetry variesamongst them. Second, a comparison could becarried out also with respect to the US. As stressedby Angeloni et al. (2003), the adjustment of euro areaoutput in the wake of a monetary policy changeappears to be primarily driven by investment changes.The evidence for the US is that monetary policyseems to have relatively stronger effect on consump-tion than on investment. This difference may be dueto a stronger dependence of euro area firms on banklending, while households in US are relatively moreindebted.
Acknowledgements
We would like to thank Alessandro Calza, DavidMarques Ibanez and Alessandro Secchi for helpfulcomments. We also thank the participants at theBank of Italy’s Regional Economics seminar and atthe 2007 ESEM conference. The usual disclaimerapplies. This article was in great part written whileLeonardo Gambacorta was affiliated with theEconomic Outlook and Monetary PolicyDepartment of the Bank of Italy. The opinionsexpressed in this article are those of the authors onlyand do not necessarily represent the views of theBank of Italy or the Bank for InternationalSettlements.
References
Alesina, A., Blanchard, O., Galı, J., Giavazzi, F. andUhlig, H. (2001) Defining a macroeconomic frame-work for the euro area, Monitoring theEuropean Central Bank 3, Centre for EconomicPolicy Research, London.
Altunbas, Y., Fazylov, O. and Molyneux, P. (2002)Evidence of a bank-lending channel in Europe,Journal of Banking and Finance, 26, 2093–110.
Angeloni, I., Kashyap, A. K. and Mojon, B. (2003)Monetary Policy Transmission in the Euro Area,Cambridge University Press, Cambridge.
Atanasova, C. (2003) Credit market imperfectionsand the business cycle dynamics: a nonlinearapproach, Studies in Nonlinear Dynamics andEconometrics, 7, 1–19.
Balke, N. (2000) Credit and economic activity: creditregimes and non-linear propagation of shocks, TheReview of Economics and Statistics, 82, 344–9.
Bernanke, B. (2006) Monetary aggregates and monetarypolicy at the Federal Reserve: a historical perspective,remarks at The Fourth ECB Central BankingConference, Frankfurt, Germany.
Bernanke, B. and Blinder, A. S. (1988) Is it money or credit,or both or neither? Credit, money and aggregatedemand, The American Economic Review, 78, 435–9.
Bernanke, B. and Gertler, M. (1989) Agency costs, networth, and business fluctuations, The AmericanEconomic Review, 79, 14–31.
Bernanke, B., Gertler, M. and Gilchrist, S. (1996) Thefinancial accelerator and the flight to quality, TheReview of Economics and Statistics, 72, 14–31.
Blinder, A. (1987) Credit rationing and effective supplyfailures, Economic Journal, 97, 327–52.
Calza, A., Manrique, M. and Sousa, J. (2006) Credit in theeuro area: an empirical investigation using aggregatedata, The Quarterly Review of Economics and Finance,46, 211–26.
Calza, A. and Sousa, J. (2006) Output and inflationresponses to credit shocks: are there threshold effectsin the euro area?, Studies in Nonlinear Dynamics andEconometrics, 10, 1–21.
Calza, A. and Zaghini, A. (2006) Non-linear dynamics inthe euro area demand for M1, ECB Working PaperNo. 592.
Casolaro, L., Eramo, G. and Gambacorta, L. (2006) Unmodello econometrico per il credito bancario alleimprese in Italia, Moneta e Credito, 59, 151–83.
Casolaro, L. and Gambacorta, L. (2005) Un modelloeconometrico per il credito bancario alle famiglie inItalia, Moneta e Credito, 58, 29–56.
Cook, S., Holly, S. and Turner, P. (1999) The power of testsfor non-linearity: the case of Granger-Lee asymmetry,Economics Letters, 62, 155–9.
De Long, J. B. and Summers, L. (1988) How doesmacroeconomic policy affect output, Brooking Paperson Economic Activity, 2, 433–80.
Ehrmann, M., Gambacorta, L., Martinez Pages, J.,Sevestre, P. and Worms, A. (2003) The effects ofmonetary policy in the euro area, Oxford Review ofEconomic Policy, 19, 58–72.
European Central Bank (ECB) (2004) Recent developmentsin loans to non-financial corporations, MonthlyBulletin, June.
European Central Bank (ECB) (2006) Longer-termdevelopments in loans to the private sector, MonthlyBulletin, May.
Freixas, X. and Rochet, J. (1997) Microeconomics ofBanking, MIT Press, Cambridge.
Friedman, B. and Kuttner, K. (1993) Economic activityand the short-term credit markets: an analysis of prices
Modelling bank lending in the euro area 1111
Dow
nloa
ded
by [
"Uni
vers
ity a
t Buf
falo
Lib
rari
es"]
at 1
1:38
07
Oct
ober
201
4
and quantities, Brooking Papers on Economic Activity,1, 193–283.
Gambacorta, L. (2005) Inside the bank lending channel,European Economic Review, 49, 1737–59.
Gambacorta, L. and Iannotti, S. (2007) Are thereasymmetries in the response of bank interest rates tomonetary shocks?, Applied Economics, 39, 2503–17.
Granger, C. and Lee, T. (1989) Investigation ofproduction, sales and inventory relationships usingmulticointegration and non-symmetric error correc-tion models, Journal of Applied Econometrics, 4,145–59.
Huelsewig, O. (2003) Bank behavior, interest rate targetingand monetary policy transmission, WuerzburgEconomic Paper No. 43.
Johansen, S. (1988) Statistical analysis of cointegrationvectors, Journal of Economic Dynamics and Control,12, 231–54.
Johansen, S. (1995) Likelihood-based inference in cointe-grated vector autoregressive models, Oxford UniversityPress, Oxford.
Johansen, S. and Nielsen, B. (1993) Asymptotics forcointegration rank tests in the presence of interventiondummies – manual for the simulation program disco,Manuscript, Institute of Mathematical Statistics,University of Copenhagen.
Kakes, J. (2000) Monetary Transmission in Europe: TheRole of Financial Markets and Credit, Edward Elgar,Cheltenham.
Karras, G. (1996) Why are the effects of money supplyshocks asymmetric? Convex aggregate supply or‘pushing on a string’?, Journal of Macroeconomics,18, 605–19.
Kashyap, A. K. and Stein, J. C. (1997) The role of banksin monetary policy: a survey with implications forthe European monetary union, Economic Perspectives,21, 2–18.
Kashyap, A. K., Stein, J. C. and Wilcox, D. W. (1993)Monetary policy and credit conditions: evidence fromthe composition of external finance, The AmericanEconomic Review, 83, 78–98.
Lang, W. and Nakamura, L. C. (1995) Flight to quality inbank lending and economic activity, Journal ofMonetary Economics, 36, 145–64.
Lim, G. C. (2001) Bank interest rate adjustments: are theyasymmetric?, The Economic Record, 77, 135–47.
McCallum, J. (1991) Credit rationing and the monetarytransmission mechanism, The American EconomicReview, 81, 946–51.
Melitz, J. and Pardue, M. (1973) The demand and supply ofcommercial bank loans, Journal of Money, Credit andBanking, 5, 669–92.
Mishkin, F. S. (1995) Symposium on the monetarytransmission mechanism, Journal of EconomicPerspectives, 9, 3–10.
Nicoletti-Altimari, S. (2001) Does money lead inflation inthe euro area?, ECB Working Paper No. 63.
Oliner, S. D. and Rodebusch, G. D. (1996) Is there a broadcredit channel for monetary policy?, Economic Review,1, 3–13.
Reimers, H. (1992) Comparisons of tests for multivariatecointegration, Statistical Papers, 33, 335–59.
Rogoff, K. (1989) Reputation, coordination, and monetarypolicy, in Modern Business Cycle Theory (Ed.)R. Barro, Harvard University Press, Cambridge,MA, pp. 236–64.
Romer, C. D. and Romer, D. H. (1990) New evidence onthe monetary transmission mechanism, BrookingPaper on Economic Activity, 1, 149–213.
Saikkonen, P. (1992) Estimation and testing of cointegratedsystems by an autoregressive approximation,Econometric Theory, 8, 1–27.
Stiglitz, J. and Greenwald, B. C. (2003) Towards a NewParadigm in Monetary Economics, CambridgeUniversity Press, Cambridge.
Stiglitz, J. and Weiss, A. (1981) Credit rationing in marketswith imperfect information, The American EconomicReview, 71, 393–410.
Stock, J. H. and Watson, M. W. (1999) Forecastinginflation, Journal of Monetary Economics, 44, 293–335.
Taylor, J. B. (1993) Discretion versus policy rules inpractice, Carnegie-Rochester Conference Series onPublic Policy, 39, 195–214.
Terasvirta, T., Tjøstheim, D. and Granger, C. W. J. (1994)Aspects of modeling non linear time series,in Handbook of Econometrics (Eds) R. F. Engle andD. L. McFadden, Vol. 4, Elsevier Science, Amsterdam,pp. 2919–37.
1112 L. Gambacorta and C. Rossi
Dow
nloa
ded
by [
"Uni
vers
ity a
t Buf
falo
Lib
rari
es"]
at 1
1:38
07
Oct
ober
201
4
