1

ARTICLE IN PRESS
European Economic Review 52 (2008) 792819 www.elsevier.com/locate/eer
How do banks set interest rates?

Leonardo Gambacorta
Research Department Banca dItalia, Via Nazionale, 91 00184 Rome, Italy Received 14 July 2006; accepted 12 June 2007 Available online 20 June 2007
Abstract This paper studies cross-sectional differences in banks interest rates. It adds to the literature in two ways. First, it analyzes systematically the micro and macroeconomic factors that inuence the pricesetting behaviour of banks. Second, by using banks prices (rather than quantities) it provides an alternative way of disentangling loan supply from loan demand shift in the bank lending channel literature. The results suggest that heterogeneity in the banking rates pass-through depending on liquidity, capitalization and relationship lending exists only in the short run. r 2007 Elsevier B.V. All rights reserved.
JEL classication: E44; E51; E52 Keywords: Monetary policy transmission; Interest rates; Bank lending channel
1. Introduction This paper studies cross-sectional differences in the price-setting behaviour of Italian banks. The study was motivated by two major considerations. First, heterogeneity in the response of bank interest rates to market rates helps to explain how monetary policy decisions are transmitted through the economy independently of the consequences for bank lending. The analysis of heterogeneous behaviour in banks interest setting has been largely neglected by the existing literature. This seems odd because, in practice, when bank prices change, real effects on consumption and investment could be produced even if there were no changes in total lending. In the US only one quarter of mortgages are at variable rates; however all mortgages can be easily renanced with virtually no cost and this implies
Tel.: +390647922979; fax: +390647923723.
E-mail address: leonardo.gambacorta@bancaditalia.it 0014-2921/$ - see front matter r 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.euroecorev.2007.06.022
ARTICLE IN PRESS
L. Gambacorta / European Economic Review 52 (2008) 792819 793
that in the case of a monetary easing the service of households debt may be signicantly reduced. In the euro area, around three quarters of household debt for housing nance is at variable rates (this percentage increases to 85 for lending to non-nancial corporations). In the UK, basically no loans are granted at rates that are xed for over 5 years. The scant evidence on the effects of monetary shocks on bank prices is mainly due to the lack of long series of micro data and contrasts with the general interest in understanding the effective adjustment of retail rates to changes in money market conditions. Second, this paper aims to add to the bank lending channel literature by identifying loan supply shocks via banks prices (rather than quantities). To date the identication problem has been solved by claiming that certain bank-specic characteristics (i.e. size, liquidity, capitalization) inuence only loan supply movements, while banks loan demand is independent of them. After a monetary tightening, the drop in the supply of credit should be larger for small banks, which are nanced almost exclusively from deposits and equity (Kashyap and Stein, 1995), less liquid banks, which cannot protect their loan portfolio against monetary tightening simply by drawing down cash and securities (Stein, 1998; Kashyap and Stein, 2000), and poorly capitalized banks, which have less access to markets for uninsured funding (Peek and Rosengren, 1995; Kishan and Opiela, 2000; Van den Heuvel, 2002).1 The intuition of an identication via prices of a loan supply shift is very simple: If loan demand is not perfectly elastic, the effect of a monetary tightening on bank interest rates should be more pronounced for small, low-liquid and low-capitalized banks. This paper adds to the existing literature also by considering additional control variables to dealt with endogeneity problems between demand and supply shifts. In particular, I consider the effects on deposit and loan demand of ination, permanent and transitory movements in real GDP, local shocks in economic activity. I introduce bank-specic measures for market power and credit risk that changes through time. The analysis nally controls for the presence of a bank capital channel, that works via the assets and liabilities mismatch in banks balance sheet, and for the existence of internal capital markets. Banks behaviour is analyzed by using a balanced panel of 73 Italian banks that represent more than 70% of the banking system. Heterogeneity is investigated with respect to the interest rate on short-term lending and that on current accounts. The use of microeconomic data is particularly appropriate in this context because aggregation may signicantly bias the estimation of dynamic economic relations (Harvey, 1981). Moreover, information at the level of individual banks provides a more precise understanding of their behavioural patterns and should be less prone to structural changes, such as the creation of EMU. I do not investigate asymmetries in the passthrough between policy tightening and easing because Gambacorta and Iannotti (2007) show that these asymmetries vanished in Italy after the introduction of the 1993 Consolidated Law on Banking. The paper is organized as follows. Section 2 describes some institutional characteristics that help to explain the behaviour of banking rates in Italy in the last two decades. Section 3 reviews the main channels that inuence banks interest rate settings by trying to
1 All these studies on cross-sectional differences in the effectiveness of the bank lending channel refer to the US. The literature on European countries is far from conclusive (see Altunbas et al., 2002; Ehrmann et al., 2003). For Italy, see Gambacorta (2005).
ARTICLE IN PRESS
794 L. Gambacorta / European Economic Review 52 (2008) 792819
disentangle macro from microeconomic factors. After a description of the econometric model and the data in Section 4, Section 5 shows the empirical results. Robustness checks are presented in Section 6. The last section summarizes the main conclusions. 2. Some facts regarding bank interest rates in Italy Before discussing the main channels that inuence banks price setting, it is important to analyze the institutional characteristics that have inuenced bank interest rates in Italy in the last two decades. There is evidence that in the 1980s, Italian banks were comparatively slow to adjust their rates (Banca dItalia, 1986; Cottarelli and Kourelis, 1994), but important measures to liberalize the markets and introduce deregulation in the last 20 years should have inuenced the speed at which changes in money market conditions are transmitted to lending and deposit rates (Cottarelli et al., 1995; Passacantando, 1996; Angelini and Cetorelli, 2002). In fact, between the mid-1980s and the early 1990s all the restrictions that characterized the Italian banking system (lending ceilings, foreign exchange controls, limits on branching and supply of long-term lending) were gradually removed. To avoid criticism of structural breaks, the econometric analysis is based on the period after the introduction of the Consolidated Law on Banking (1993:032001:03), when all the main reforms of the Italian banking system had already taken place. The behaviour of bank interest rates in Italy reveals some stylized facts (Figs. 1 and 2). First, there has been a substantial fall in average rates since the end of 1992, and second there has been a strong and persistent dispersion of rates among banks. These stylized facts suggest that both the time series and the cross-sections dimensions are important for understanding the banks interest setting. This justies the use of panel data techniques. The main reason for the fall in bank interest rates is the EMU related convergence of ination and of the monetary policy stance. As a result, the interbank rate decreased by more than 10 percentage points in the period 199399. Excluding the episode of turbulence on the foreign exchange markets in 1995, it moved upwards from the third quarter of 1999
20 Period before C.L.B. 1987:01-1993:02 15 Estimation period 1993:03-2001:03
10
Euro
Short term lending rate Repo rate 3-month interbank rate Interest rate on current accounts
0 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Fig. 1. Banking interest rates (quarterly data, percentage points).
ARTICLE IN PRESS
0.350 Interest rate on short-term loans 0.300 0.250 0.200 0.150 0.100 0.050 0.000 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 0.350 Interest rate on current accounts 0.300 0.250 0.200 0.150 0.100 0.050 0.000 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 ACROSS BANKS OVER TIME
Fig. 2. Cross-sectional and time series dispersion of interest rates. (a) Interest rate on short-term loans. (b) Interest rate on current accounts. The coefcient of variation is given by the ratio of the standard errors to the mean. The series that refers to the variability over time (dotted line) shows the coefcient of variation worked out in each year on the 12 average monthly gures (obtained on the whole dataset). By contrast, the series that capture the variability across banks (solid line) shows the coefcient of variation of annual averages of bankspecic interest rates. In this case, rst I work out the average annual interest rate for each bank and than I calculate the coefcient of variation to this 73 data in all years.
ARTICLE IN PRESS
to the end of 2000, then continued its declining trend. From a statistical point of view, this behaviour calls for the investigation of a possible structural break in the 1990s.2 The second stylized fact is cross-sectional dispersion among interest rates. Fig. 2 shows the coefcient of variation for loan and deposit rates both over time and across banks in the period 19872001. The temporal variation (dotted line) of the two rates shows a different behaviour from the mid-1990s when the deposit rate is more variable, probably for a catching-up process of the rate toward a new equilibrium caused by the convergence process. Moreover, the cross-sectional dispersion of the deposit rate is greater than that of the loan rate, especially after the introduction of the euro.3 3. What does inuence banks interest rate setting? The literature that studies banks interest rate setting behaviour generally assumes that banks operate under oligopolistic market conditions (Santomero, 1984). This means that a bank does not act as a price-taker but sets its loan rates taking into account the demand for loans and deposits. This section reviews the main channels that inuence banks interest rates (all symbols are summarized in Table 1). 3.1. Loan and deposit demand The interest rate on loans depends positively on real GDP and ination (y and p). Better economic conditions increase the number of projects becoming protable in terms of expected net present value and hence increase the demand for credit (Kashyap et al., 1993). As stressed by Melitz and Pardue (1973) only boosts in permanent income (yP) have a positive inuence on loan demand, while the effect due to the transitory part (yT) could also be associated with a self-nancing effect that reduces the proportion of bank debt (Friedman and Kuttner, 1993).4 An increase in the money market rate (iM) raises the opportunity cost of other forms of nancing (i.e. bonds), making lending more attractive. This mechanism also boosts loan demand and increases the interest rate on loans. The interest rate on deposits is negatively inuenced by real GDP and ination. A higher level of income increases the demand for deposits and therefore reduces the incentive for banks to set higher deposit rates. In this case, the shift of deposit demand will be higher if the transitory component of GDP is affected (unexpected income is generally rst deposited in current accounts). On the contrary, an increase in the money market rate, Ceteris paribus, makes it more attractive to invest in risk-free securities that represent an
In the period 199598, which coincides with the convergence process towards stage three of EMU, it will be necessary to allow for a change in the statistical properties of interest rates (see Gambacorta and Iannotti, 2007). 3 In the period before the 1993 Consolidated Law on Banking deposit interest rates were quite sticky to monetary policy changes. Deposit interest rate rigidity in this period has been extensively analyzed for the US as well. Among the market factors that have been found to affect the responsiveness of bank deposit rates are the direction of the change in market rates (Hannan and Berger, 1991), whether the bank interest rate is above or below a target rate (Hutchison, 1995; Neumark and Sharpe, 1992), and market concentration in the banks deposit market (Hannan and Berger, 1991). Rosen (2002) develops a model of price settings in the presence of heterogeneous customers to explain why bank deposit interest rates respond sluggishly to some extended movements in money market rates but not to others. 4 Taking this into account, in Section 4 I tried to disentangle the two effects using a Beveridge and Nelson (1981) decomposition. More details are provided in the Appendix.
2
ARTICLE IN PRESS
L. Gambacorta / European Economic Review 52 (2008) 792819 Table 1 Variables description Variables Dependent variables Fixed effects Macro variables Symbols iL iD mk iM yP, yT p Xk Description Interest rate on domestic short-term loans Interest rate on current account deposits Bank-specic dummy variable Monetary policy indicator Permanent and transitory components of real GDP computed using the Beveridge and Nelson (1981) decomposition Ination rate Size: Log of total assets Kashyap and Stein (1995), Ehrmann et al. (2003) Liquidity: Cash and securities over total assets Stein (1998), Kashyap and Stein (2000) Excess capital: Difference between regulatory capital and capital requirements Peek and Rosengren (1995), Kishan and Opiela (2000), Gambacorta and Mistrulli (2004) Deposit strength: Ratio between deposits and bonds plus deposits Berlin and Mester (1999), Weth (2002) Credit relationship: Ratio between long term loans and total loans Berger and Udell (1992) Bank-specic cost of monetary policy due to maturity transformation. This variable captures interest rate risk Ratio between bad loans and total loans. This variable captures the riskiness of lending operations and should be offset by a higher expected yield on loans Management efciency: Ratio of total loans and deposits to the number of branches (in alternative cost-to-total asset ratio) Interest rate volatility: Coefcient of variation of iM Herndahl index in local markets Convergence dummy: Step dummy that takes the value of 1 in the period 1995:031998:03 and 0 elsewhere Seasonal dummies 797
Bank-specic characteristics that inuence the bank lending channel
Bank capital channel Credit risk
ck jk
Efciency ratio Interest rate volatility Market structure Dummies
ek s H F
Note: For more information on the denition of the variables, see the Appendix.
alternative to detaining deposits; the subsequent reduction in deposit demand determines an upward pressure on the interest rate on deposits. 3.2. Bank efciency, credit risk and interest rate volatility The costs of intermediation (screening, monitoring, branching costs, etc.) have a positive effect on the interest rate on loans and a negative effect on that on deposits (efciency is represented by e). The interest rate on lending also depends on the riskiness of the credit portfolio; banks that invest in riskier projects will have a higher rate of return in order to compensate the higher percentage of bad loans that have to be written off ( j). Bank interest rates are also inuenced by interest rate volatility. High volatility of the money market rate (s) should increase lending and deposit rates. Following the dealership model by Ho and Saunders (1981) and its extension by Angbazo (1997), the interest rate on loans should be more affected by interbank interest rate than that on
ARTICLE IN PRESS
deposits (diL/ds4diD/ds). This should indicate a positive correlation between interest rate volatility and the spread. 3.3. Interest rate channel Bank interest rates are also directly inuenced by monetary policy changes. A monetary tightening (easing) determines a reduction (increase) in reservable deposits and an increase (reduction) in market interest rates. This has a positive effect on bank interest rates through the traditional interest rate channel. Nevertheless, the increase in the cost of nancing could have a different impact on banks depending on their specic characteristics. There are two channels through which heterogeneity among banks may produce a different impact on lending and deposit rates: The bank lending channel and the bank capital channel. Both mechanisms are based on adverse selection problems that affect banks fund-raising, but from different perspectives. 3.4. Bank lending channel According to the bank lending channel thesis, a monetary tightening has an effect on bank loans because the drop in reservable deposits cannot be completely offset by issuing other forms of funding (i.e. uninsured CDs or bonds; for an opposite view see Romer and Romer, 1990) or liquidating some assets. Kashyap and Stein (1995, 2000), Stein (1998) and Kishan and Opiela (2000) claim that the market for bank debt is imperfect. Since nonreservable liabilities are not insured and there is an asymmetric information problem about the value of banks assets, a lemons premium is paid to investors. According to these authors, small, low-liquid and low-capitalized banks pay a higher premium because the market perceives them to be more risky. Since these banks are more exposed to asymmetric information problems they have less capacity to shield their credit relationships in the case of a monetary tightening, and they should cut their supplied loans and raise their interest rates by a larger amount. Moreover, these banks have less capacity to issue bonds and CDs and therefore they could try to contain the drain of deposits by raising their rates more. Two aspects should be stressed. First, to avoid endogeneity problems bank-specic characteristics should refer to the period before banks set their interest rates. Second, heterogeneous effects, if any, should be detected only in the short run, while there is no a priori reason why these effects should inuence the long-run relationship between interest rates. Apart from the standard indicators of size (logarithm of total assets), liquidity (cash and securities over total assets) and capitalization (excess capital over total assets),5 two other bank-specic characteristics are worth investigating: (a) The ratio between deposits and bonds plus deposits; (b) the ratio between long-term loans and total loans. The rst indicator may be named deposit strength. Banks that have a large amount of deposits will adjust their deposit rates by less (and less quickly) than banks whose liabilities
5 It is important to note that the effect of bank capital on the bank lending channel cannot be easily captured by the capital-to-asset ratio. This measure, generally used by the existing literature to analyze the distributional effects of bank capitalization on lending, does not take into account the riskiness of a bank portfolio. A relevant measure is instead the excess capital that is the amount of capital banks hold in excess of the minimum required to meet prudential regulation standards. For more details, see Gambacorta and Mistrulli (2004).
ARTICLE IN PRESS
are mainly composed by variable rate bonds that are directly affected by market movements (Berlin and Mester, 1999). The intuition of this result is that, given the presence of menu costs, it is more likely that a bank will adjust its terms for passive deposits if the conditions of its own alternative form of renancing (i.e. bonds) change. Moreover a bank will refrain from changing conditions on deposits because, if their proportion on total liabilities is high, even small changes to their price will have huge effect on total interest rate costs. On the contrary banks which use relatively more bonds than deposits for nancing purposes fall under greater pressure because their costs increase contemporaneously and to similar extent to market rates. Therefore, an important indicator in analyzing the pass-through between market and banking rates is the ratio between deposits and bonds plus deposits. In particular, Berlin and Mester (1999) claim that a solid deposit base should generate a smoother loan pricing too. The ratio between long-term loans and total loans, consistently with Berger and Udell (1992), represents a proxy for long-term business. Since Italian banks typically extend both long-term and short-term loans to the same borrower, high values of the ratio, other things being equal, should be associated with those credit institutions that maintain close ties with their non-bank customers. These banks should adjust their short-term lending rates comparatively less and slowly: They may offer implicit interest rate insurance to riskaverse borrowers in the form of below-market rates during periods of high market rates, for which the banks are later compensated when market rates are low. For example, Weth (2002) nds that in Germany those banks with large volumes of long-term business with households and rms change their short-term interest rates less frequently than the others. 3.5. Bank capital channel The bank capital channel is based on the fact that bank assets typically have a longer maturity than liabilities (Van den Heuvel, 2002). After an increase in market interest rates, a smaller fraction of loans can be renegotiated with respect to deposits (loans are mainly long term, while deposits are typically short term): Banks therefore incur a cost due to the maturity mismatch that reduces prots and then capital accumulation. If equity is sufciently low and it is too costly to issue new shares, banks reduce lending (otherwise they fail to meet regulatory capital requirements) and widen their interest rate spread. This leads to an increase in the interest rates on loans and a decrease in those on deposits: In the oligopolistic version of the MontiKlein model, the maturity transformation cost has the same effect as an increase in operating costs. The explicit inclusion of the assets and liabilities mismatch (calculated taking into account derivatives) among the determinants of bank interest rate setting represents one of the novelty of this paper (see Stein (2002) for a discussion). 3.6. Industry structure The literature underlines two possible impacts of concentration on the pricing behaviour of banks (Berger and Hannan, 1989). A rst class of model claims that more concentrated banking industry will behave oligopolistically (structureperformance hypothesis), while another class of model stresses that concentration is due to more efcient banks taking over less efcient counterparts (efcientstructure hypothesis). This means that in the rst case lower competition should result in higher spreads, while in the second case a decrease
ARTICLE IN PRESS
in managerial costs due to increased efciency should have a negative impact on the spread. In the empirical part, great care will be paid therefore to market structure effects by introducing a specic measure of the degree of competition that each bank faces (H). Nevertheless, the scope of this paper is not to extract policy implications for this issue, for which a different analysis is needed (see Focarelli and Panetta, 2003). 4. Empirical specication and data The empirical specication used in this paper adapts the standard approach for the estimation of bank rates to the case of heterogeneous banks. Following Cottarelli et al. (1995), Lim (2000) and Weth (2002) we start from two simple error correction models that establish a long-run relationship between each bank rate and the money market rate. Economic theory on oligopolistic (and perfect) competition suggests that, in the long run, both bank rates (on lending and deposits) should be related to the level of the monetary rate that reects the marginal yield of a risk-free investment (Klein, 1971).6 Panel cointegration between interest rates have been preliminarily checked by means of the methodology discussed in Pedroni (1999) that allows heterogeneity in both the cointegrating vector as well as in the dynamics associated with short-run deviations from the long-run equilibrium: All tests overwhelming rejected the null of no cointegration between the interest rates. The starting point of the analysis was therefore the two following ECM equations7:
DiL k;t mLk
2 X j1
kLj DiL k;tj
2 X j0
bLj DiM tj aL iL k;t1 gL iM t1
2 X j0
fLj Z tj GL Fk;t L k;t ,
(1)
DiD
k;t
mDk
2 X j1
kDj DiD
k;tj
2 X j0
bDj DiM
tj
aD i D
k;t1
gD iM
t1
2 X j0
fDj Ztj GD Fk;t D k;t
(2) with k 1,y,N (k banks) and t 1,y,T (t periods). Data are quarterly (1993:032001:03) and not seasonally adjusted. The subscript L stands loans, D for deposits. The panel is balanced with N 73 banks. Two lags have been selected in order to obtain well behaved residuals. The description of the variables is reported in Table 1. The vector Z D ln yP ; D ln yT ; p; Dck ; j k ; ek ; s includes stationary variables that inuence interest rates in the short run8; F is a vector of dummies. The model allows for xed effects across banks, as
6 Freixas and Rochet (1997) show that in a model of imperfect competition among N banks, if a part of deposits (Z) is invested in compulsory reserves, the long-run relationships among lending, deposit and money market rates become: iL iM+mark-up and iD (1Z)iM+mark-down. 7 A simple model that justies the choice of the specication is reported in the working paper version of the study (see http://www.nber.org/papers/W10295). In that model the risk neutral bank, that operates under oligopolistic market conditions, set interest rates, taking into account a loan and a deposit demand. Solving the maximization problem the optimal level of the interest rates are a function not only of the given money market rate but also of the macroeconomic variables that inuence loan and deposit demand and of bank-specic characteristics. Lending and deposit quantities wash out in the calculation and do not enter the structural equations. 8 The possible presence of unit roots have been investigated by means of the Levin et al. (2002) test for the bankspecic variables (Dck, jk, ek) and the Augmented Dickey Fuller (ADF) test for the time series (Dln yP, Dln yT, p, s). All the variables in vector Z resulted stationary.
ARTICLE IN PRESS
indicated by the bank-specic intercept mk. The long-run pass-through between each bank rate and the money market rate is given by g/a, while the loading coefcient is represented by a. Asymmetric effects across banks due to bank-specic characteristics (size, liquidity, capitalization, deposit strength, long-run relationship) are analyzed following the approach used by Kashyap and Stein (2000) and by Ehrmann et al. (2003). In particular, Eqs. (1) and (2) are modied by introducing interaction terms between interest rates and the bank-specic characteristics Xm (with m 1,y,5) that capture heterogeneity in the monetary transmission mechanism. The bank-specic characteristics are also introduced alone to control for distributional effects in interest rate changes independent of monetary policy. We have:
DiL
k;t
mLk
2 X j1
kLj DiL
k;tj
2 X j0
bLj
5 X m1
! b X m;k;t1 DiM Lmj

5 X m1
tj
5 X m1
lLm X m;k;t1
aL
5 X m1
! g X m;k;t1 iM Lm
t1
a X m;k;t1 iL Lm
k;t1
gL
2 X j0
fLj Z tj GL Fk;t L k;t ,
3
DiD
k;t
mDk
2 X j1
kDj DiD
k;tj
2 X
bDj
5 X m1
! b X m;k;t1 DiM Dmj

5 X m1
tj
5 X m1
lDm X m;k;t1
fDj Ztj GD Fk;t D k;t ,
aD
5 X m1
j0
! g X m;k;t1 iM Dm
t1
a X m;k;t1 iDk;t1 Dm
gD
2 X j0
4 where the bank-specic characteristics X refers to t1 to avoid an endogeneity bias. The long-run pass-through between each bank rate and the money market rate is given P P by g n g X m;k;t1 a n a X m;k;t1 . Therefore to test if the pass-through m1 m1 between the money market rate and the bank rate is complete it is necessary to verify that this coefcient is equal to one. If this is the case there is a one-to-one long-run relationship between the lending (deposit) rate and the money market rate, while the individual effect mkinuences the bank-specic mark-up (mark-down). The loading coefcient P a n a X m;k;t1 must be signicantly negative if the assumption of an equilibrium m1 relationship is correct. In fact, it represents what percentage of an exogenous variation from the steady state between the rates is brought back towards equilibrium in the next period.10 The degree of bank interest rate stickiness in the short run can be analyzed by the impact multiplier and the total effect after 3 months. The variable Xm,k,t1 represents a bank-specic characteristic that economic theory suggests inuences only loan and deposit supply movements, without affecting loan and
9 Given the complexity of the model, we have followed a general to specic strategy in order to drop some statistically insignicant variables in the Z vector. Nevertheless, this approach has not be interpreted as a mechanical reduction process that implies dropping all insignicant parameters (Pagan, 1990): The restrictions have been tested comparing the initial model with the reduced one. In the nal models, only contemporaneous exogenous variables are included with the exception, as expected, of the risk measure for loans jk in the equation for interest rate deposits. Moreover, the coefcients b2 and b2* were never signicant. The estimated equations are reported on the bottom of Tables 35. 10 Testing for heterogeneity in the loading coefcient means verifying if a* is signicant or not. At the same time heterogeneity in the long-run pass-through can be proved if a*gag* is statistically different from zero.
ARTICLE IN PRESS
deposit demand. In particular, all bank-specic indicators (wm,k,t1) have been reparameterized in the following way: X m;k;t1 wm;k;t
T X t1
PN
k1 wm;k;t N
!, T
Each indicator is therefore normalized with respect to the average across all the banks in the respective sample in order to obtain a variable whose sum over all observations is zero.11 This has two implications. First, the interaction terms between interest rates and Xm,k,t1 in Eqs. (3) and (4) are zero for the average bank (this because X m;k;t1 0). Second, the coefcients bj, a and g are directly interpretable as average effects. 4.1. Characteristics of the dataset The dataset includes 73 banks that represent more than 70% of the total Italian banking system in terms of loans over the whole sample period. Since information on interest rates is not available for mutual banks, the sample is biased towards large banks. Foreign banks and special credit institutions are also excluded. This bias toward large banks has two consequences. First, the distributional effects of the size variable must be treated with extreme caution because a small bank inside this sample should not be considered to have the same characteristics as where the full population of Italian banks is used.12 The size grouping in this study mainly controls for variations in scale, technology and scope efciencies across banks but it is not able to shed light on differences between mutual and other banks. Second, results for the average bank will provide more macroeconomic insights than studies based on the whole population (where the average bank size is very small). Table 2 gives some basic information on the dataset. Rows are organized dividing the sample with respect to the bank-specic characteristics that are potential causes of heterogeneous shifts in loan supply in the event of changes in monetary policy. In the columns, the table reports summary statistics for the two interest rates and for each indicator. Several clear patterns emerge. Considering size, small banks charge higher interest rates on lending but show a lower time variation. This ts the standard idea of a close customer relationship between small rms and small banks that on the one hand, determines a rent extraction (Petersen and Rajan, 1994) and, on the other, provides them with an incentive to smooth the effect of a monetary tightening (Angelini et al., 1998). Moreover, small banks are more liquid and capitalized than average, and this should help them to reduce the effect of cyclical variations on supplied credit. On the liability side, the percentage of deposits (overnight deposits, CDs and savings accounts) is greater among small banks, while their bond issues are more limited than those of large banks. Nevertheless, no
The size indicator has been normalized with respect to the mean on each single period. This procedure removes trends in size (for more details see Ehrmann et al., 2003). 12 In particular, banks that are considered small in this study are labeled medium in other studies of the Italian banking system that analyze quantities (see Gambacorta and Mistrulli, 2004; Gambacorta, 2005). This is clear if one considers that the average assets of a small bank in my data (1.6 billion euros) over the sample period are very similar to those of a medium bank in the whole system (1.7 billion euros).
11
Table 2 Summary statistics (1993:032001:03) Bank characteristics (*) Number of Interest rate on short-term lending banks Mean St. dev. Min Max 73 18 18 18 18 18 18 9.51 9.28 10.02 9.51 9.33 9.71 9.42 11.78 7.77 8.51 10.97 2.72 2.81 2.73 2.72 2.73 2.73 2.81 1.49 2.24 2.59 2.12 3.69 3.69 5.03 3.69 4.42 3.69 4.75 4.88 3.69 3.69 4.00 16.12 15.06 16.12 15.94 14.86 16.12 15.93 16.12 15.06 15.06 16.12 Interest rate on current accounts Mean 3.58 3.57 3.55 3.57 3.61 3.68 3.53 5.15 2.41 2.80 4.68 St. dev. 1.79 1.74 1.79 1.80 1.71 1.80 1.79 0.96 1.45 1.67 1.44 Min 0.52 0.73 0.52 0.65 0.73 0.52 0.74 0.74 0.52 0.65 0.53 Max 8.21 7.35 8.21 8.21 7.35 7.18 8.21 8.21 7.35 7.36 7.43 16.20 51.15 1.55 4.67 43.75 9.66 24.28 6.58 27.00 21.92 8.51 24.00 19.01 25.11 33.07 14.91 26.15 20.82 29.69 18.56 19.98 28.26 3.91 2.56 4.81 4.27 3.13 6.86 1.49 4.46 3.42 3.80 3.95 82.40 77.60 84.40 86.27 72.43 85.49 78.40 98.53 66.10 71.84 93.13 37.66 38.98 41.72 36.15 43.66 37.22 38.46 28.72 45.30 53.29 22.46 Size (1) Liq. (2) Cap. (3) DS (4) CR (5) L. Gambacorta / European Economic Review 52 (2008) 792819
Total sample (1) Big banks Small banks (2) Liquid banks Low-liquid banks (3) Well-capitalized banks Low-capitalized banks
ARTICLE IN PRESS
(4) Banks with high DS ratio 18 Banks with low DS ratio 18 (5) Banks with high CR ratio 18 Banks with low CR ratio 18
Former special credit institutions, foreign banks and banche di credito cooperativo are excluded. The sample represents more than 70% of the total system in terms of lending. All interest rates are annualized and given in percentages. (1) The size indicator is given by total asset (billions of euros). (2) The liquidity indicator is represented by the sum of cash and government securities over total assets. (3) The capital ratio is given by excess capital divided by total assets. Excess capital is the difference between regulatory capital and total capital requirements. (4) The deposit strength indicator (DS) is the ratio between deposits and deposits plus bonds. (5) The credit relationship indicator (CR) is the ratio between long-term loans and total loans. (*) A bank with a high characteristic has an average ratio above the rst quartile of the distribution. A bank with a low characteristic has an average ratio below the third quartile. Since the characteristics of each bank could change with time, percentiles have been worked out on mean values. For more details regarding the denition of the variables, see the Appendix. The sources of the dataset are Bank of Italy supervisory returns and 10-day reports.
803
ARTICLE IN PRESS
signicant differences emerge in the level and volatility of the interest rate on current accounts. High-liquid banks are smaller than average and are more capitalized. These characteristics should reduce the speed of the bank lending channel transmission through interest rates. In particular, since deposits represent a large share of their funding they should have a smoother transmission on the rate of current accounts. Well-capitalized banks make relatively more short-term loans. In general they are not listed and issue less subordinated debt to meet capital requirements. This evidence is consistent with the view that, Ceteris paribus, capitalization is higher for those banks that bear more adjustment costs from issuing new (regulatory) capital. Well-capitalized banks charge a higher interest rate on lending; this probably depends on their higher ratios of bad loans, which increase their credit risk. In other words, their higher capitalization is necessary to cope with a riskier portfolio. Moreover, the interest rate on deposits is lower for low-capitalized banks, indicating that agents do not perceive these deposits as riskier than those at other banks. This has two main explanations. First, the impact of bank failures has been very small in Italy, especially with respect to deposits.13 Second, the presence of deposit insurance insulates the deposits of less-capitalized banks from the risk of default. The deposit strength and the long-run relationship indicators seem to have a high ability to explain heterogeneity in banks price setting behaviour. Differences in the standard deviations of the two groups are particularly sensitive, calling for lower interest rate variability of banks with a high percentage of deposits and long-term loans. 5. Results The main channels that inuence the interest rate on short-term lending and that on current accounts are summarized, respectively, in Tables 3 and 4. The rst part of each table shows the inuence of the permanent and transitory components of real GDP and ination. These macro variables capture cyclical movements and serve to isolate shifts in loan and deposit demand from monetary policy changes. The second part of the tables presents the effects of bank efciency, credit and interest rate volatility. The third part highlights the effects of the interest rate channel (the monetary policy effect for the average bank). These are divided into four components: (i) The immediate pass-through; (ii) the one-quarter pass-through; (iii) the long-run pass-through between each bank rate and the monetary policy indicator; (iv) the loading coefcient of the cointegrating relationship. Table 5 shows test for the existence of distributional effects produced by bank-specic characteristics in the monetary transmission mechanism.14 The last two parts of Tables 3
During our sample period, the share of deposits of failed banks to total deposits approached 1% only twice, namely in 1987 and 1996 (Boccuzzi, 1998). 14 The immediate pass-through is given by the expression b0+b0*Xk,t1 and heterogeneity among banks is simply tested through the signicance of b0*. The effect for a bank with a low value of the characteristic under 0:25 0:25 evaluation is worked out through the expression b0 b X k;t1 , where X k;t1 is the average for the banks below 0 0:75 the rst quartile. Vice versa the effect for a bank with a high value of the characteristic is calculated using X k;t1 . The total effect after three months for the average bank is given by b0(1+a1+k1)+b1+g0 , while heterogeneity among banks can be accepted if and only if the expression b0 a b 1 a k1 b g X k;t1 a b X 2 0 1 0 k;t1 is equal to zero. The long-run pass-through is given by: g g X k;t1 =a a X k;t1 , while the loading coefcient is a a X k;t1 . Standard errors have been approximated with the delta method (Rao, 1973).
13
L. Gambacorta / European Economic Review 52 (2008) 792819
Table 3 Results for the equation on the interest rate on short-term lending
Dependent variable: Quarterly change of the interest rate on short-term lending (I) Benchmark model (II) Alternative efciency indicator Coeff. S.E. (III) Herndhal index (IV) Bank-specic control for loan demand shifts Coeff. S.E.
ARTICLE IN PRESS
Coeff. (1) Loan demand Ination (jL) Permanent income (dL1) Transitory income (dL2) Bank-specic regional GDP (dL) (2) Efciency, credit risk and int. rate vol. Bank efciency 1 (xL, units of interm. funds per branch) Bad loans (yL) Interest rate volatility (cL) Bank efciency 2 (x0 L, cost-to-total asset ratio) (3) Interest rate channel (effects for the average bank) Immediate pass-through (bL0) Pass-through after a quarter (bL0(1+aL1+kL1)+bL1+gL) Long run pass-through (gL/aL) Test for unitary long run pass-through (H0: gL/aL 1) Loading of the long-run relationship (aL)
S.E.
Coeff.
S.E.
0.155*** 0.032*** 0.014
0.017 0.012 0.025
0.150*** 0.033*** 0.012
0.015 0.012 0.026
0.145*** 0.047*** 0.018
0.015 0.012 0.026
0.123***
0.035
0.042*
0.023
0.003* 0.018*** 0.011***
0.002 0.002 0.001
0.017*** 0.010*** 0.047***
0.001 0.001 0.018
0.007*** 0.019*** 0.013***
0.002 0.001 0.001
0.009** 0.018*** 0.013***
0.004 0.003 0.001
0.452*** 0.879*** 1.004***
0.062 0.039 0.015 0.790 0.050
0.465*** 0.884*** 1.010***
0.064 0.045 0.017 0.556 0.056
0.441*** 0.820*** 1.008***
0.058 0.048 0.016 0.619 0.047
0.455*** 0.880*** 1.020***
0.059 0.040 0.020 0.317 0.046
0.354***
0.360***
0.400***
0.420***
805
806
Table 3 (continued ) Dependent variable: Quarterly change of the interest rate on short-term lending (I) Benchmark model (II) Alternative efciency indicator Coeff. S.E. (III) Herndhal index (IV) Bank-specic control for loan demand shifts Coeff. S.E.
Coeff. (4) Bank capital channel Interest rate risk (ZL) (5) Market structure Herndahl index in local credit markets Miss-specication tests MA(1), MA(2) (p-value) Sargan test (p-value) No of banks, no of observations 0.000 73
S.E.
Coeff.
S.E.
0.196***
0.062
0.187***
0.056
0.158** 0.147*
0.068 0.080 0.933 0.981 2336
0.203** 0.091* 0.000 73
0.098 0.051 0.424 0.986 2336
0.255 0.886 2336
0.000 73
0.305 0.808 2336
0.000 73
ARTICLE IN PRESS
This table shows the results of the Eq. (3) for the interest rate on short-term lending
2 X j1 1 X j0 5 X m1 5 X m1
! b X m;k;t1 DiM Lmj

5 X m1 tj
DiL
k;t
mLk
kLj DiL
k;tj
bLj
5 X m1
lLm X m;k;t1 dL1 D ln yP dL2 D ln yT t t
! g X m;k;t1 iM Lm
aL
a X m;k;t1 iL Lm
k;t1
gL
t1
jL pt ZL Dck;t1 yL j k;t xL ek;t cL st GL Fk;t Lk;t with k 1,y,N (k number of banks) and t 1,y,T (t periods). Data are quarterly (1993:032001:03) and not seasonally adjusted. The ve bank-specic characteristics are size, liquidity, capitalization, deposit strenght (the ratio between deposits and bonds plus deposits), long-run relationship (the ratio between longterm and total loans). The panel is balanced with N 73 banks. The description of the variables is reported in Table 1. The models have been estimated using the GMM estimator suggested by Arellano and Bond (1991), which ensures efciency and consistency provided that the models are not subject to serial correlation of order two and that the instruments used are valid (which is tested for with the Sargan test). * signicance at the 10% level; ** Idem, 5%; *** Idem, 1%.
Table 4 Results for the equation on interest rate on current accounts

Dependent variable: Quarterly change of the interest rate on current accounts (I) Benchmark model (II) Alternative efciency indicator Coeff. S.E. (III) Herndhal index (IV) Bank-specic control for deposit demand shifts Coeff. S.E.
ARTICLE IN PRESS
Coeff. (1) Deposit demand Ination (jD) Permanent income (dD1) Transitory income (dD2) Bank-specic regional GDP (dD) (2) Efciency, credit risk and int. rate vol. Bank efciency 1 (xD, units of interm. funds per branch) Bad loans (yD) Interest rate volatility (cD) Bank efciency 2 (x0 D, cost-to-total asset ratio) (3) Interest rate channel (effects for the average bank) Immediate pass-through (bD0) Pass-through after a quarter (bD0(1+aD1+kD1)+bD1+gL): Long run pass-through (gD/aD): Test for unitary long run pass-through (H0: gD/aD 1) Loading of the long-run relationship (aD)
S.E.
Coeff.
S.E.
0.066*** 0.050*** 0.196***
0.012 0.007 0.011
0.061*** 0.058*** 0.221***
0.012 0.005 0.011
0.051*** 0.046** 0.180***
0.012 0.022 0.011
0.022*
0.013
0.055*
0.032
0.002* 0.000 0.002***
0.001 0.002 0.001
0.000 0.002*** 0.010***
0.002 0.001 0.005
0.002* 0.000 0.007***
0.001 0.003 0.002
0.004** 0.000 0.008***
0.002 0.003 0.001
0.450*** 0.545*** 0.696*** 0.570***
0.042 0.033 0.010 0.000 0.043
0.445*** 0.544*** 0.698*** 0.580***
0.044 0.026 0.018 0.000 0.044
0.450*** 0.550*** 0.700*** 0.531***
0.053 0.044 0.020 0.000 0.042
0.455*** 0.557 0.701*** 0.551***
0.038 1.038 0.020 0.000 0.042
807
808
Table 4 (continued ) Dependent variable: Quarterly change of the interest rate on current accounts (I) Benchmark model (II) Alternative efciency indicator Coeff. S.E. (III) Herndhal index (IV) Bank-specic control for deposit demand shifts Coeff. S.E.
Coeff. (4) Bank capital channel Interest rate risk (ZD) (5) Market structure Herndahl index Miss-specication tests MA(1), MA(2) (p-value) Sargan test (p-value) No. of banks, no. of observations 0.000 73
S.E.
Coeff.
S.E.
0.044***
0.012
0.049***
0.010
0.074*** 0.137*
0.014 0.072 0.444 0.573 2336
0.075*** 0.110* 0.000 73
0.016 0.061 0.230 0.900 2336
0.878 0.410 2336
0.000 73
0.785 0.494 2336
0.000 73
ARTICLE IN PRESS
This table shows the results of the Eq. (4) for the interest rate on current accounts
2 X j1 1 X j0 5 X m1 5 X m1
! b X m;k;t1 DiM Dmj

5 X m1 tj
DiD
k;t
mDk
kDj DiD
k;tj
bDj
5 X m1
lDm X m;k;t1 dD1 D ln yP dD2 D ln yT t t
! g X m;k;t1 iM Dm
aD
a X m;k;t1 iD Dm
k;t1
gD
t1
jD pt ZD Dck;t1 yD j k;t xD ek;t cD st GD Fk;t Dk;t with k 1,y, N (k number of banks) and t 1,y,T (t periods). Data are quarterly (1993:032001:03) and not seasonally adjusted. The ve bank-specic characteristics are size, liquidity, capitalization, deposit strenght (the ratio between deposits and bonds plus deposits), long-run relationship (the ratio between longterm and total loans). The panel is balanced with N 73 banks. The description of the variables is reported in Table 1. The models have been estimated using the GMM estimator suggested by Arellano and Bond (1991), which ensures efciency and consistency provided that the models are not subject to serial correlation of order two and that the instruments used are valid (which is tested for with the Sargan test). * signicance at the 10% level; ** Idem, 5%; *** Idem, 1%.
Table 5 Bank lending channel (1) Size Coeff. S.E. (2) Liquidity Coeff. S.E. (3) Capitalization Coeff. S.E. (4) Deposit strength Coeff. S.E. (5) LR relationship L. Gambacorta / European Economic Review 52 (2008) 792819 Coeff. S.E.
(A) Dependent variable is the quarterly change of the interest rate on short-term lending Immediate pass-through Ho: No heterogeneity (p-value) 0.159 0.027 Low characteristic 0.492*** 0.064 0.476*** 0.058 0.558*** High characteristic 0.393*** 0.080 0.421*** 0.069 0.308*** Pass-through after a quarter Ho: No heterogeneity (p-value) 0.639 0.317 Low characteristic 0.895*** 0.058 0.891*** 0.036 0.914*** High characteristic 0.857*** 0.040 0.868*** 0.033 0.847*** Long run pass-through Ho: No heterogeneity (p-value) 0.511 0.813 Low characteristic 0.997*** 0.014 1.002*** 0.015 1.017*** High characteristic 1.046*** 0.016 1.019*** 0.015 1.001*** Loading of the LR relationship Ho: No heterogeneity (p-value) 0.681 0.990 Low characteristic 0.377*** 0.072 0.354*** 0.063 0.318*** High characteristic 0.324*** 0.092 0.354*** 0.046 0.399*** (B) Dependent variable is the quarterly change of the interest rate on current accounts Immediate pass-through Ho: No heterogeneity (p-value) 0.972 0.129 Low characteristic 0.453*** 0.043 0.470*** 0.050 0.479*** High characteristic 0.452*** 0.050 0.434*** 0.037 0.419*** Pass-through after a quarter Ho: No heterogeneity (p-value) 0.160 0.978 Low characteristic 0.572*** 0.032 0.546*** 0.038 0.566*** High characteristic 0.524*** 0.043 0.545*** 0.033 0.517***
0.043 0.065 0.110 0.744 0.082 0.075 0.884 0.013 0.012 0.536 0.070 0.095
0.460*** 0.437***
0.702 0.059 0.077 0.913 0.050 0.047 0.930 0.026 0.012 0.761 0.089 0.085
0.519*** 0.375***
0.016 0.050 0.084 0.912 0.039 0.053 0.615 0.016 0.023 0.773 0.086 0.095
ARTICLE IN PRESS
0.883*** 0.876***
0.888*** 0.873***
1.016*** 1.004***
0.986*** 1.034***
0.332*** 0.375***
0.332*** 0.376***
0.529 0.054 0.074 0.481 0.055 0.045
0.509*** 0.400***
0.032 0.044 0.054 0.203 0.045 0.039
0.497*** 0.406***
0.112 0.062 0.062 0.224 0.041 0.034
0.590*** 0.516***
0.563*** 0.525***
809
810
Table 5 (continued ) (1) Size Coeff. Long run pass-through Ho: No heterogeneity (p-value) Low characteristic High characteristic Loading of the LR relationship Ho: No heterogeneity (p-value) Low characteristic High characteristic S.E. (2) Liquidity Coeff. S.E. (3) Capitalization Coeff. S.E. (4) Deposit strength Coeff. S.E. (5) LR relationship Coeff. S.E.
0.703*** 0.697***
0.881 0.014 0.013 0.388 0.048 0.074
0.685*** 0.714***
0.205 0.010 0.009 0.820 0.050 0.047
0.688*** 0.719***
0.467 0.009 0.011 0.481 0.019 0.025
0.688*** 0.711***
0.442 0.006 0.009 0.004 0.062 0.054
0.706*** 0.692***
0.722 0.011 0.011 0.575 0.059 0.051
0.537*** 0.612***
0.565*** 0.575***
0.607*** 0.523***
0.452*** 0.680***
0.589*** 0.550***
ARTICLE IN PRESS
This table shows heterogeneity tests for the benchmark model (column I in Tables 3 and 4). Panel A refers to Eq. (3) for the interest rate on short-term lending. Panel B presents results for Eq. (4) on current accounts. The model is given by:
2 X j1 1 X 5 X m1
! b X m;k;t1 Wmj
5 X m1
DiW
k;t
mWk
kWj DiW
5 X m1
k;tj
bWj
DiM
tj
5 X m1
lWm X m;k;t1 dW 1 D ln yP dW 2 D ln yT t t
j0
! g X m;k;t1 iM Wm
aW
a X m;k;t1 iW Wm
k;t1
gW
t1
jW pt ZW Dck;t1 yW j k;t xW ek;t cW st GW Fk;t Wk;t with iw quarterly change of the interest rate on short-term lending (w L) or current accounts (w D), k 1,y,N (k number of banks) and t 1,y,T (t periods). Data are quarterly (1993:032001:03) and not seasonally adjusted. The ve bank-specic characteristics are size, liquidity, capitalization, deposit strength (the ratio between deposits and bonds plus deposits), long-run relationship (the ratio between long-term and total loans). The panel is balanced with N 73 banks. The description of the variables is reported in Table 1. The models have been estimated using the GMM estimator suggested by Arellano and Bond (1991) which ensures efciency and consistency provided that the models are not subject to serial correlation of order two and that the instruments used are valid (which is tested for with the Sargan test). A bank with low characteristic has an average ratio below the rst quartile, a bank with high characterisic has an average ratio above third quartile. * signicance at the 10% level; ** Idem, 5%; *** Idem, 1%.
ARTICLE IN PRESS
and 4 analyze the effects determined by the bank capital channel and by the market structure. The models have been estimated using the GMM estimator suggested by Arellano and Bond (1991), which ensures efciency and consistency provided the models are not subject to serial correlation of order two and the instruments used are valid (which is tested for with the Sargan test).15 The use of instrumental variables is crucial to avoid endogeneity problems; for example, some Z variables in Eqs. (3) and (4), such as permanent income and ination, could determine not only loan demand but also current policy rates. 5.1. Loan and deposit demand As predicted by theory, only changes in permanent income have a positive and signicant effect on the interest rate on short-term lending, while the transitory component is never signicant (see column I of Table 3). In fact, as discussed in Section 3, the effect of transitory changes in our benchmark equation may also be due to a self-nancing effect that reduces the proportion of bank debt. On the contrary, the interest rate on deposits is negatively inuenced by real GDP (see column I of Table 4). In this case, the effect is higher when a change in the transitory component occurs because it is directly channelled through current accounts. The effect of ination is positive on both interest rates but is signicantly higher for short-term lending. 5.2. Bank efciency, credit risk and interest rate volatility Bank efciency reduces the interest rate on loans and increases that of deposits. Nevertheless, the effect is marginally signicant at conventional levels. These results called for an immediate robustness check using a different efciency indicator, the cost-to-total asset ratio. This indicator turned out to be more signicant than the other one, leaving all other results unchanged (see column II of Tables 3 and 4). The relative amount of bad loans has a positive and signicant effect on the interest rate on loans. This is in line with the standard result that banks that invest in riskier projects ask for a higher rate of return to compensate the credit risk. Both bank rates are positively correlated with money market rate volatility. The correlation is higher for the interest rate on loans than for that on deposits. This is consistent with the prediction of the dealership model by Ho and Saunders (1981) and its extension by Angbazo (1997), where an increase in interbank interest rate volatility is associated with a higher spread. 5.3. Interest rate channel A monetary tightening positively inuences bank interest rates. After a 1% increase in the monetary policy indicator, interest rates on short-term lending is immediately raised by around 0.5%, and by around 0.9% after a quarter. Moreover, the pass-through is complete in the long run (the null hypothesis of a unitary coefcient is not rejected in all
15 In the GMM estimation, instruments are the second lag of the dependent variable and of the bank-specic characteristics included in each equation. Ination, GDP growth rate and the monetary policy indicator are considered exogenous variables.
ARTICLE IN PRESS
models). The reaction of the short-term lending rate is greater than in previous studies of the Italian case and this calls for an increase in competition after the introduction of the 1993 Consolidated Law on Banking. Cottarelli et al. (1995), analyzing the period 19861993, nd that the immediate pass-through is around 0.2%, while the effect after 3 months is 0.6%. Their long-run pass-through is equal to 0.9%, but also in their model the null hypothesis of a complete pass-through is not rejected. The long-run pass-through of the interest rate on current accounts is 0.7%. This result is in line with the recent ndings by de Bondt et al. (2005) and Gambacorta and Iannotti (2007) under a similar sample period and a little higher than the coefcient in Angeloni et al. (1995) for the period 19871993. The hypothesis of a complete pass-through is rejected. The loading coefcients are signicantly negative at around 0.4 in the loan equation and 0.6, in the current account equation. This means that if an exogenous shock occurs, respectively, 40% and 60% of the deviation is cancelled out within the rst quarter in each bank rate. 5.4. Bank lending channel Table 5 summarizes results for heterogeneity in the bank interest rate pass-through among intermediaries. It is organized in two panels (A and B) that refer, respectively, to Eq. (3) for the interest rate on short-term lending and Eq. (4) for the interest rate on current accounts. There are also ve columns, one for each of the bank-specic characteristics under investigation (size, liquidity, capitalization, deposit strength and long-run relationship). As before the monetary policy effects are divided into four components: (i) The immediate pass-through; (ii) the one-quarter pass-through; (iii) the long-run pass-through between each bank rate and the monetary policy indicator; (iv) the loading coefcient of the cointegrating relationship. For each moment of the pass-through a test of heterogeneity in the response of the bank is provided (Ho: No heterogeneity). All results are also represented graphically in Fig. 3. From a quick look at Table 5 and Fig. 3, it emerges clearly that no heterogeneity emerges in the long-run relationship between each bank rate and the monetary policy indicator. Differences are present only in the short run and vanish after 3 months (I have highlighted in bold all those cases that reject the presence of homogenous behaviour among banks). As regards the interest rate on short-term lending, liquid and well-capitalized banks react less to a monetary policy shock. Also banks with a high proportion of long-term lending tend to change their prices less. Size is not signicant. This evidence matches previous results in the literature. Liquid banks can protect their loan portfolio against a monetary tightening simply by drawing down cash and securities (Stein, 1998; Kashyap and Stein, 2000). Well-capitalized banks that are perceived as less risky by the market are better able to raise uninsured funds in order to compensate the drop in deposits (Peek and Rosengren, 1995; Kishan and Opiela, 2000; Van den Heuvel, 2002; Gambacorta and Mistrulli, 2004). Therefore, the effects on lending detected for liquid and well-capitalized banks are mirrored by their higher capacity to insulate clients from the effects on interest rates as well. It is interesting to note that, in contrast with the evidence for the US (Kashyap and Stein, 1995), the interaction terms between size and monetary policy are insignicant. The fact that the interest rate on the short-term lending of smaller banks is
ARTICLE IN PRESS
A
1.2
Effects on the interest rate on short-term lending of a 1 percent increase in the money market rate Average Bank 1.2
1.00 0.88 BIG SMALL 0.86 0.90
Size
1.00 1.00
Liquidity 1.2 0.8

HIGH LIQUID LOW LIQUID 0.87 0.89 1.001.00
0.8
0.45
0.8
0.49
0.4 0.0 0 1.2 0.8 0.4 0.0 0 3 LR

HIGH CAPITAL LOW CAPITAL 0.85 0.91 0.56 0.31
0.4 0.0 3 Capitalization

1.00 1.00
0.39
0.4 0.0 0 3 Deposit strength LR 1.2 0.8
0.42
0.48
LR 1.2 0.8 0.4 0.0
LR
Long-run relationship
HIGH BU LOW BU 1.00 1.00 0.870.89
HIGH DEP LOW DEP
1.00 1.00 0.88 0.88
0.44 0.46
0.52
0.4 0.0
0.38
LR
LR
B
0.8
Effects on the interest rate on current accounts of a 1 percent increase in the money market rate Average Bank
0.70 0.55 0.45
Size 0.8
BIG SMALL 0.45 0.45 0.70 0.70 0.52 0.57
Liquidity 0.8
HIGH LIQUID LOW LIQUID 0.43 0.47 0.70 0.70 0.55 0.55
0.4
0.4
0.4
0.0 0 0.8 3 Capitalization

HIGH CAPITAL LOW CAPITAL 0.42 0.48 0.52 0.70 0.70 0.57
0.0 LR 0.8 0 3 LR Deposit strength

0.70 0.70 0.59 0.52
0.0 0 0.8 3 LR Long-run relationship

HIGH BU LOW BU 0.50 0.53 0.41 0.70 0.70 0.56
HIGH DEP LOW DEP 0.51 0.40
0.4
0.4
0.4
0.0 0 3 LR
0.0 0 3 LR
0.0 0 3 LR
Fig. 3. Bank lending channel effects. (A) Effects on the interest rate on short-term lending of a 1% increase in the money market rate (B) Effects on the interest rate on current accounts of a 1% increase in the money market rate. Note: The presence of signicant distributional effects is highlighted. P-values are provided in Table 5.
not more sensitive to monetary policy than that of larger banks is well documented in the literature for Italy and reects the close customer relationship between small banks and small rms (Angeloni et al., 1995; Angelini et al., 1998). This result is also consistent with Ehrmann et al. (2003), where size does not emerge as a useful indicator for the distributional effect of monetary policy on lending, not only in Italy but also in France, Germany and Spain. As regards the interest rate on current accounts, the deposit strength indicator is the only bank-specic characteristic that explains heterogeneity in banks price-setting behaviour. In particular, banks that have a high proportion of deposits raised on local
ARTICLE IN PRESS
markets (whose price is directly chosen by the banks depending on their market power) will adjust their interest rate on current accounts by less (and less quickly) than banks that heavily depend on bonds, typically issued on international market (whose price reacts directly to money market movements). In other words, as explained in Section 3, other things being equal, it is more likely that a bank will adjust its terms on deposits if the other conditions of renancing change. The liability structure seems to inuence not only the short-run adjustment but also the loading coefcient. This implies that banks with a high proportion of deposits react less when there is a deviation in the long-run mark-down: These banks have more room for adjusting their prices towards the optimal equilibrium. As expected, no cross-sectional differences emerge among banks due to size, liquidity and capitalization because current accounts are typically insured. This is consistent with the ndings for current accounts in Gambacorta (2005). 5.5. Bank capital channel As expected the maturity mismatch between bank assets and liabilities has a positive effect on the interest rate on short-term lending and a negative effect on the interest rate on current accounts. This is consistent with the existence of a bank capital channel (as well as with the bank lending channel, see Section 3). The absolute values of the coefcients are greater in the rst case, calling for a stronger adjustment on credit contracts than on deposits. Since this channel can be interpreted similarly to a general increase in the costs for banks, it is worth comparing this result with that obtained for the efciency indicator. In both cases the effect is strongest for the interest rate on short-term lending, and this is consistent with the view that the interest rate on deposits is more sluggish. 6. Other robustness checks The robustness of the results has been checked in several ways. The rst test was to introduce as additional control variable a bank-specic measure of the degree of competition that each bank faces in the market. In particular, the average value of the Herndahl index in the different local markets (corresponding to the administrative provinces of Italy) in which the bank operates was introduced in each equation (see column III in Tables 3 and 4). The reason for this test is that the xed effect remains stable over the whole period, while the degree of competition in loan and deposit markets could change over time due to the effect of concentration. The Herndahl index appears to be marginally signicant while the results of the study did not change. The second test wanted to check if demand controls (permanent and transitory GDP growth and ination) were sufcient enough to insulate supply shifts. Indeed, the absence of some demand characteristics may limit the capacity to insulate genuine supply shifts. Let us consider for example two banks that operate in two markets with a different growth of economic activity (low and high). In this case an increase in market interest rates from a monetary tightening will raise the cost of funds for both banks, but the probability of default will be likely to be higher for the bank that operate in a context of low economic growth. In order to capture this aspect, I have weighted GDP growth at the province level with the effective volume of activity that each bank has at the local level (for loans and deposits). In other words, I have considered in each equation a bank-specic GDP growth that is dependent on the evolution of the economic activity of the local markets in which
ARTICLE IN PRESS
each bank effectively operate. Since seasonality may be different across provinces I used seasonally adjusted GDP series. The results reported in column IV of Tables 3 and 4 show that no signicant changed occurred (also the interaction terms were virtually unchanged, leaving the distributional effects reported in Table 5 and Fig. 3 unaltered). The third test was to consider whether different scal treatments over the sample period could have changed deposit demand (from June 1996 the interest rate on current accounts is subject to 27% tax, deducted at source; 12.5% before). However, when the net interest rate on current accounts was used in place of gross rate, nothing changed. The fourth robustness check was the introduction of a dummy variable to take account of the spike in the change of the repo interest rate caused by the turbulence in the foreign exchange market in the rst quarter of 1995. The results are the same in this case too. The nal robustness check was to introduce a dummy variable that indicates whether the bank belongs to a group (1) or not (0). Banks belonging to a group may be less inuenced by monetary changes if they can benet from internal liquidity management; in other words, bank holding companies establish internal capital markets in an attempt to allocate capital among their various subsidiaries (Houston and James, 1998; Upper and Worms, 2004). The introduction of this dummy did not change the results of the study. 7. Conclusions This paper investigates the factors that inuence the price-setting behaviour of Italian banks. It adds to the existing literature in two ways. First, it analyzes systematically a wide range of micro and macroeconomic variables that have an effect on bank interest rates: Permanent and transitory changes in income, interest and credit risk, interest rate volatility, bank efciency. Second, the analysis of bank prices (rather than quantities) provides an alternative way of disentangling loan supply from loan demand shift in the bank lending channel literature. The main results of the study are the following. First, heterogeneity in the bank rate pass-through exists, but it is detected only in the short run: No differences exist in the longrun elasticities of bank rates to the money market rate. Second, consistently with the existing literature, interest rates on short-term lending of liquid and well-capitalized banks react less to changes in ofcial rates. Also banks with a high proportion of long-term lending tend to modify their prices less. Heterogeneity in the pass-through on the interest rate on current accounts depends on banks liability structure. Bank size is never relevant. These results are broadly consistent with the bank lending channel hypothesis. Even though the paper is not giving an answer to all possible identication issues, its aim is to improve our knowledge of the monetary transmission mechanism, also by considering its results in light of the existing literature on quantities. For example, one possible interpretation of the result by Kashyap and Stein (2000) that less liquid banks cut their volume lending by more following a monetary tightening is that this could be driven by loan demand factors (i.e. illiquid banks could systematically have more cyclical borrowers) instead than a shift in supply. If this were true, we should observe a signicant drop in loan demand following the tightening and a consequent smaller increase in bank interest rates. However, this is not true on the base of our evidence and this favours a supply-based interpretation. The results in this paper complement those in Gambacorta (2005), that analyzes the existence of a bank lending channel in Italy using data on quantities. The only difference is
ARTICLE IN PRESS
that in that paper the heterogeneous effects on lending stocks are detected also in the long run, while in this paper those on bank interest rates vanish after a quarter. This difference between the quantity and the price effect depends on the fact that while interest rates are linked in the long run by a unique mark-up relationship, the stock of credit adjusts differently in the long run depending on specic bank balance-sheet characteristics. In other words, in the case of a persistent 1% increase of the monetary policy indicator, bank interest rates augment with a different velocity towards the new equilibrium level, while the stock of credit is reduced to counterbalance the persistent drop in deposit. At the individual bank level, this drop in supplied lending will depend on bank balance sheet characteristics. For example, using the semi-elasticities calculated in Gambacorta (2005; column II in Table 2) in the long run the supply of credit, Ceteris paribus, is reduced by 0.4% by banks with a high liquidity buffer, that are able to protect their portfolio simply by drawing down cash and securities, and by 1.3% by low liquid banks. The policy implication of these combined results is that monitoring bank-specic characteristics is relevant both for evaluating the overall effects on lending and deposits and the consequences on bank rates, at least at the beginning of the monetary policy transmission mechanism. For example, in the case of a monetary tightening, if the banking system is characterized by a low level of liquidity remaining in line with the example above the monetary authority should expect a wider effect on consumption and investment not only because the reduction in supplied lending is more severe but also because bank interest rates will move more rapidly to meet the new level of ofcial rates with additional consequences on aggregate demand. Acknowledgements This study was developed while the author was a visiting scholar at the NBER. I wish to thank two anonymous referees and Juergen von Hagen for very helpful comments. I am also grateful to participants at seminars held at the Banca dItalia, the Swiss National Bank, the Bundesbank and the Banque de France. The opinions expressed in this paper are those of the author only and in no way involve the responsibility of the Bank of Italy and the NBER. Appendix. Technical details regarding the data The dataset has been constructed using three sources. Interest rates are taken from the 10-day survey conducted by the Bank of Italy. Information on bank balance sheets comes from the Banking Supervision Register at the Bank of Italy. Data on macroeconomic variables are taken from International Financial Statistics. Data on interest rates refer to transactions in euros (Italian lira before 1999). The deposit interest rate is the weighted average rate paid by the single banks on current accounts, which are highly homogenous deposit products. The rate on domestic short-term lending for the single bank is the weighted average of all lending positions. Data on overdraft and other fees are not available and are therefore excluded from the computation (this could be an issue because banks in principle could change these conditions instead of or in addition to their interest rates). The choice of the short-term rate as a measure of the bank interest lending pass-through is based on a number of considerations. First, shortterm lending excludes subsidized credit. Second, short-term loans typically are not
ARTICLE IN PRESS
collateralized and this allows the bank lending channel to be insulated from the balance sheet channel. Broadly speaking, the pass-through from market interest rates to the interest rate on loans does not depend on market price variations that inuence the value of collateral. Both interest rates are posted rates that are changed at discrete intervals, often less frequently than weekly (Green, 1998). Since a substantial part of short-term loans (more than 60%) in Italy is represented by current account overdrafts and advances against trade credit receivable, whose interest rates are revised automatically on a monthly basis, the quarterly frequency of the data (I calculated averages of 10-day gures) should be sufcient to capture these price changes. Both bank rates are gross of scal deductions. The interest rate taken as monetary policy indicator is that on repurchase agreements between the Bank of Italy and credit institutions in the period 199398, and the interest rates on main renancing operations of the ECB for the period 19992001. The series does not present any break. The variable ck,t rk,t1DiMt represents the bank-specic cost of monetary policy due to maturity transformation. In particular, rk,t1 measures the loss per unit of asset a bank suffers when the monetary policy interest rate is raised by 1%. The cost at time t is inuenced by the maturity transformation in t1. This variable is computed according to supervisory regulations on interest rate risk exposure that depends on the maturity mismatch among assets and liabilities. To work out the real cost, we have therefore multiplied rk,t1 by the change that has occurred in interest rates. Therefore, ck,t rk,t1DiMt represents the cost (gain) that a bank suffers (obtains) in each quarter and this measure inuences the level of bank interest rates. Since the model is expressed in error correction form we have included this variable in rst differences (other details are provided in Gambacorta and Mistrulli, 2004). In assembling our sample, the so-called special credit institutions (long-term credit banks) have been excluded as they were subject to different supervisory regulations regarding the maturity range of their assets and liabilities. Nevertheless, special long-term credit sections of commercial banks have been considered part of the banks to which they belonged. Particular attention has been paid to the treatment of mergers. In practice, it has been assumed that these took place at the beginning of the sample period, summing the balancesheet items of the merging parties. For example, if bank A was incorporated by bank B at time t, bank B has been reconstructed backwards as the sum of the merging banks before the merger. Bank interest rates have been reconstructed backwards using as weights the short-term loans and current accounts of the merging parties. The same methodology has been used, among others by Peek and Rosengren (1995), Kishan and Opiela (2000) and Ehrmann et al. (2003). Only banks reporting detailed lending and deposit rates over the whole sample period were considered. I refrain from adopting short-time series to ensure sufcient asymptotic in the context of the error correction estimation. Bank observations that were missing or misreported or that constituted clear outliers were excluded from the sample. Bad loans are dened as loans for which legal proceedings have been instituted to obtain repayment. The permanent component of GDP has been computed using the Beveridge and Nelson (1981) decomposition. An ARIMA model (1,1,1) was applied to the logarithm of the series. Computations have been carried out using the algorithm described in Newbold (1990). Robustness of the results have been checked by means of a statistical analysis of the residuals.
ARTICLE IN PRESS
The possible presence of structural breaks in interest rate series has been investigated using the procedure developed by Banerjee et al. (1992) based on sequential test for changes in the mean of each interest rate series. The hypothesis of this procedure is that, if there is a break, its date is not known a priori but rather is gleaned from the data. The results in Gambacorta and Iannotti (2007) show clearly that unit-root/no-break null can be rejected at the 2.5% critical value level against the stationarity/mean-shift alternative for the period 1995:031998:03. In all the equations a convergence dummy, that takes the value of 1 in this period and 0 elsewhere, has been introduced. References
Altunbas, Y., Fazylow, O., Molyneux, P., 2002. Evidence on the bank lending channel in Europe. Journal of Banking and Finance 26 (11), 20932110. Angbazo, L., 1997. Commercial bank net interest margins, default risk, interest-rate risk, and off-balance sheet banking. Journal of Banking and Finance 21 (1), 5587. Angelini, P., Cetorelli, N., 2002. The effects of regulatory reform on competition in the banking industry. Journal of Money, Credit and Banking 35 (5), 663684. Angelini, P., Di Salvo, P., Ferri, G., 1998. Availability and cost of credit for small businesses: customer relationships and credit cooperatives. Journal of Banking and Finance 22 (68), 925954. Angeloni, I., Buttiglione, L., Ferri, G., Gaiotti, E., 1995. The credit channel of monetary policy across heterogeneous banks: The case of Italy. Banca dItalia, Temi di discussione, No. 256. Arellano, M., Bond, S., 1991. Some tests of specication for panel data: Monte Carlo evidence and an application to employment equations. Review of Economic Studies 58, 277297. Banca dItalia, 1986. Modello trimestrale delleconomia italiana. Banca dItalia, Temi di discussione, No. 80. Banerjee, A., Lumsdaine, R.L., Stock, J.H., 1992. Recursive and sequential tests of the unit-root and trend break hypotheses: Theory and international evidence. Journal of Business and Economic Statistics 10 (3), 271287. Berger, A.N., Hannan, T.H., 1989. The price-concentration relationship in banking. Review of Economics and Statistics 71, 291299. Berger, A.N., Udell, G.F., 1992. Some evidence on the empirical signicance of credit rationing. Journal of Political Economy 100 (5), 10471077. Berlin, M., Mester, L.J., 1999. Deposits and relationship lending. Review of Financial Studies 12 (3), 579607. Beveridge, S., Nelson, C., 1981. A new approach to the decomposition of economics time series into permanent and transitory components with particular attention to measurement of the business cycle. Journal of Monetary Economics 21 (2), 151174. ` Boccuzzi, G., 1998. La crisi dellimpresa bancaria. Proli economici e giuridici. Giuffre, Milano. Cottarelli, C., Kourelis, A., 1994. Financial structure, bank lending rates and the transmission mechanism of monetary policy. IMF Staff Papers 41 (4), 587623. Cottarelli, C., Ferri, G., Generale, A., 1995. Bank lending rates and nancial structure in Italy: A case study. IMF Working Papers, No. 38. de Bondt, G., Mojon, B., Valla, N., 2005. Term structure and the sluggishness of retail bank rates in the Euro area. ECB, Working Paper Series, No. 518. Ehrmann, M., Gambacorta, L., Martinez Pages, J., Sevestre, P., Worms, A., 2003. Financial systems and the role of banks in monetary policy transmission in the Euro area. In: Angeloni, I., Kashyap, A., Mojon, B. (Eds.), Monetary Policy Transmission in the Euro Area. Cambridge University Press, Cambridge. Focarelli, D., Panetta, F., 2003. Are merger benecial to consumers? Evidence from the market for bank deposits. American Economic Review 93 (4), 11521172. Freixas, X., Rochet, J., 1997. Microeconomics of Banking. MIT Press, Cambridge. Friedman, B., Kuttner, K., 1993. Economic activity and the short-term credit markets: An analysis of prices and quantities. Brooking Papers on Economic Activity 2, 193283. Gambacorta, L., 2005. Inside the bank lending channel. European Economic Review 49, 17371759. Gambacorta, L., Iannotti, S., 2007. Are there asymmetries in the response of bank interest rates to monetary shocks? Applied Economics, forthcoming. Gambacorta, L., Mistrulli, P., 2004. Does bank capital affect lending behavior? Journal of Financial Intermediation 13 (4), 436457.
ARTICLE IN PRESS
L. Gambacorta / European Economic Review 52 (2008) 792819 819 Green, C.J., 1998. Banks as interest rate managers. Journal of Financial Services Research 14 (3), 189208. Hannan, T.H., Berger, A.N., 1991. The rigidity of prices: Evidence from banking industry. American Economic Review 81 (4), 938945. Harvey, 1981. Time Series Models. Allan, Oxford. Ho, T.S.Y., Saunders, A., 1981. The determinants of bank interest margins: Theory and empirical evidence. Journal of Financial and Quantitative Analysis 16 (2), 581600. Houston, J.F., James, C., 1998. Do bank internal capital market promote lending? Journal of Banking and Finance 22 (68), 899918. Hutchison, D.E., 1995. Retail bank deposit pricing: An intertemporal asset pricing approach. Journal of Money Credit and Banking 27 (1), 217231. Kashyap, A., Stein, J.C., 1995. The impact of monetary policy on bank balance sheets. Carnegie Rochester Conference Series on Public Policy 42, 151195. Kashyap, A., Stein, J.C., 2000. What do a million observations on banks say about the transmission of monetary policy. American Economic Review 90 (3), 407428. Kashyap, A., Stein, J.C., Wilcox, D., 1993. Monetary policy and credit conditions: Evidence from the composition of external nance. American Economic Review 83 (1), 7898. Kishan, R.P., Opiela, T.P., 2000. Bank size, bank capital and the bank lending channel. Journal of Money, Credit and Banking 32 (1), 121141. Klein, M., 1971. A theory of the banking rm. Journal of Money, Credit and Banking 3 (2), 205218. Levin, A., Lin, C.F., Chu, C., 2002. Unit root tests in panel data: Asymptotic and nite-sample properties. Journal of Econometrics 108, 124. Lim, G.C., 2000. Bank interest rate adjustments: Are they asymmetric? The Economic Record 77 (237), 135147. Melitz, J., Pardue, M., 1973. The demand and supply of commercial bank loans. Journal of Money, Credit and Banking 5 (2), 669692. Neumark, D., Sharpe, S.A., 1992. Market structure and the nature of price rigidity: Evidence from the market for consumer deposits. Quarterly Journal of Economics 107 (2), 657680. Newbold, P., 1990. Precise and efcient computation of the BeveridgeNelson decomposition of economic time series. Journal of Monetary Economics 26 (3), 453457. Pagan, A.R., 1990. Three econometric methodologies: A critical appraisal. In: Granger, C.W.J. (Ed.), Modelling Economic Series: Readings in Econometric. Oxford University Press, Oxford, pp. 97120. Passacantando, F., 1996. Building an institutional framework for monetary stability. BNL Quarterly Review 49 (196), 83132. Pedroni, P., 1999. Critical values for cointegration tests in heterogeneous panels with multiple regressors. Oxford Bulletin of Economics and Statistics 61, 653670. Peek, J., Rosengren, E.S., 1995. Bank lending and the transmission of monetary policy. In: Peek, J., Rosengren, E.S., (Eds.), Is Bank Lending Important for the Transmission of Monetary Policy? Federal Reserve Bank of Boston Conference Series No. 39, pp. 4768. Petersen, M., Rajan, R., 1994. The benets of lending relationships: Evidence from small business data. The Journal of Finance 49 (1), 337. Romer, C.D., Romer, D.H., 1990. New evidence on the monetary transmission mechanism. Brooking Paper on Economic Activity 1, 149213. Rosen, R.J., 2002. What goes up must come down? Asymmetries and persistence in bank deposit rates. Journal of Financial Service Research 21 (3), 173193. Santomero, A.M., 1984. Modeling the banking rm: A survey. Journal of Money Credit and Banking 16 (4), 576602. Stein, J.C., 1998. An adverse-selection model of bank asset and liability management with implications for the transmission of monetary policy. RAND Journal of Economics 29 (3), 466486. Stein, J.C., 2002. Commentary to Does Bank Capital Matter for Monetary Transmission? by S. Van den Heuvel. FRBNY Economic Policy Review May, 267270. Upper, C., Worms, A., 2004. Estimating bilateral exposures in the German interbank market: Is there a danger of contagion? European Economic Review 48 (4), 827849. Van den Heuvel, S.J., 2002. Does bank capital matter for monetary transmission? FRBNY Economic Policy Review 8 (1 May), 259266. Weth, M.A., 2002. The pass-through from market interest rates to bank lending rates in germany. Discussion Paper No. 11, Economic Research Center of the Deutsche Bundesbank.

1

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

1

Hochgeladen von

Copyright:

Verfügbare Formate

ARTICLE IN PRESS

European Economic Review 52 (2008) 792819 www.elsevier.com/locate/eer

How do banks set interest rates?

20 Period before C.L.B. 1987:01-1993:02 15 Estimation period 1993:03-2001:03

Bank-specic characteristics that inuence the bank lending channel

Bank capital channel Credit risk

Efciency ratio Interest rate volatility Market structure Dummies

kLj DiL k;tj

bLj DiM tj aL iL k;t1 gL iM t1

fLj Z tj GL Fk;t L k;t ,

fDj Ztj GD Fk;t D k;t

! b X m;k;t1 DiM Lmj

fLj Z tj GL Fk;t L k;t ,

! b X m;k;t1 DiM Dmj

L. Gambacorta / European Economic Review 52 (2008) 792819

0.155*** 0.032*** 0.014

0.017 0.012 0.025

0.150*** 0.033*** 0.012

0.015 0.012 0.026

0.145*** 0.047*** 0.018

0.015 0.012 0.026

0.003* 0.018*** 0.011***

0.002 0.002 0.001

0.017*** 0.010*** 0.047***

0.001 0.001 0.018

0.007*** 0.019*** 0.013***

0.002 0.001 0.001

0.009** 0.018*** 0.013***

0.004 0.003 0.001

0.452*** 0.879*** 1.004***

0.062 0.039 0.015 0.790 0.050

0.465*** 0.884*** 1.010***

0.064 0.045 0.017 0.556 0.056

0.441*** 0.820*** 1.008***

0.058 0.048 0.016 0.619 0.047

0.455*** 0.880*** 1.020***

0.059 0.040 0.020 0.317 0.046

0.068 0.080 0.933 0.981 2336

0.203** 0.091* 0.000 73

0.098 0.051 0.424 0.986 2336

L. Gambacorta / European Economic Review 52 (2008) 792819

0.255 0.886 2336

0.305 0.808 2336

! b X m;k;t1 DiM Lmj

lLm X m;k;t1 dL1 D ln yP dL2 D ln yT t t

L. Gambacorta / European Economic Review 52 (2008) 792819

Table 4 Results for the equation on interest rate on current accounts

0.066*** 0.050*** 0.196***

0.012 0.007 0.011

0.061*** 0.058*** 0.221***

0.012 0.005 0.011

0.051*** 0.046** 0.180***

0.012 0.022 0.011

0.002* 0.000 0.002***

0.001 0.002 0.001

0.000 0.002*** 0.010***

0.002 0.001 0.005

0.002* 0.000 0.007***

0.001 0.003 0.002

0.004** 0.000 0.008***

0.002 0.003 0.001

0.450*** 0.545*** 0.696*** 0.570***

0.042 0.033 0.010 0.000 0.043

fLj Z tj GL Fk;t L k;t ,

fDj Ztj GD Fk;t D k;t

fLj Z tj GL Fk;t L k;t ,

0.155* 0.032* 0.014

0.150* 0.033* 0.012

0.145* 0.047* 0.018

0.003* 0.018* 0.011*

0.017* 0.010* 0.047***

0.007* 0.019* 0.013***

0.009 0.018* 0.013***

0.452* 0.879* 1.004***

0.465* 0.884* 1.010***

0.441* 0.820* 1.008***

0.455* 0.880* 1.020***

0.066* 0.050* 0.196***

0.061* 0.058* 0.221***

0.051* 0.046 0.180***

0.000 0.002* 0.010*

0.004 0.000 0.008*

0.450* 0.545* 0.696* 0.570*

0.445* 0.544* 0.698* 0.580*

0.450* 0.550* 0.700* 0.531*

0.455* 0.557 0.701* 0.551***