BANK INTEREST RATE PASS-THROUGH IN THE EURO AREA: THE IMPACT OF THE FINANCIAL CRISIS.
The study aims at an evaluation of the interest rate pass-through (IRPT) process of the Euro area member countries. This process refers to the speed and the extent to which monetary policy interest rates are transmitted to bank interest rates and it is of pivotal importance for the functioning of monetary policy transmission mechanism. More specifically, the higher the speed and the magnitude of the transmission of central bank interest rates changes into banking market interest rates, the higher the effectiveness of monetary policy. Moreover, in the context of a monetary union, the homogeneity of the IRPT process among members is a key prerequisite for the effective conduct of the single monetary policy. Empirical literature provides evidence of high degree of heterogeneity of the IRPT processes among euro area members. Moreover, the post-2008 international financial crisis produced asymmetrical effects in the financial sectors of the euro area members and, thus, the key question is to what extent such developments deteriorated the degree of homogeneity in the monetary policy transmission processes (Barigozzi, Conti, & Luciani, 2013).
Hence, the aim is to evaluate the impact of the post-2008 financial crisis on the IRPT processes within the Euro area and to examine the determinants of the pass-through variability. To this end, the authors use the recently produced data by the European Central Bank on the cost of borrowing from banks for the 2003-2013 period. According to ECB, the cost of borrowing indices are better approximants of the banking conditions in each member country and thus, they are more useful for cross-country comparisons relative to interest rates for specific banking market segments. In most studies, the IRPT is examined by regressing the relevant banking market interest rates on monetary policy or money market rates. The model used here also includes specific banking market demand and supply side variables in an attempt to evaluate their impact on the pass-through mechanism. Therefore, the authors of this study use unemployment rates to capture demand conditions and banking sector ratios to take into account supply side factors. In addition, the authors use the government bond spreads to capture the impact of country risk.
The authors use the pooled mean group estimator advanced by Pesaran, Shin, and Smith (1999) which restricts the long-run pass-through coefficients to be identical across countries allowing only the short-run dynamics to differ. Then, Hausman (1978) specification tests are used to check the validity of this assumption. The use of the Hausman test on the long-run homogeneity restriction is a way to check for the existence of a significant dispersion in the long-run relation between the cost of borrowing and the rest of the explanatory variables across countries. The acceptance of the long-run homogeneity restriction indicates that IRPT heterogeneity comes mainly from the short-run dynamics. On the contrary, the rejection of the long-run homogeneity restriction indicates that part of the overall cross-country IRPT heterogeneity is also explained by heterogeneous long-run pass-through relations.
In order to examine the effects of the post-2008 financial crisis on the pass-through mechanism, the authors estimate the model for two sub-samples, focusing both on the pre-2008 crisis and the post-crisis periods. To split the sample, the authors use the time point derived by relevant specific tests for structural breaks in the pass-through relationships.
Previous studies focused on the transmission mechanism of monetary policy rate onto lending and deposit rates. ? policy rate or a policy affected money market rate is considered as exogenous variable affecting the interest rate of the final banking products. The study by Cottarelli and Kourelis (1994) was one of the first that followed this approach; they employed an autoregressive distributed lag specification and data from 31 industrial and developing countries to assess the completeness of the IRPT process. They found significant cross-country heterogeneity, a result which is attributed to structural differences in the financial sectors. Studies that followed incorporated cointegration methodologies in analyzing IRPT. As bank interest rates and market rates are most often characterized by unit roots, one can identify the long-run cointegrating relationships in the error correction framework. Such studies, among many others, include Sander and Kleimeier (2001), Mojon (2000), Heinemann and Schuler (2002), Toolsema, Sturm and de Haan (2002), de Bondt (2002), Sander and Kleimeier (2004), and Karagiannis, Panagopoulos, and Vlamis (2010).
Sander and Kleimeier (2001) examined the IRPT process within the euro area and provided evidence that not only the speed of adjustment differed but also the nature of the adjustment process itself was heterogeneous across member countries. Mojon (2000) incorporated additional structural indicators, balance sheet, and competition indices as additional determinants in the pass-through process. The author found that national asymmetries within the euro area remained substantial in each of the banking market segments examined. Karagiannis, Panagopoulos, and Vlamis (2010) examined the IRPT process in the USA and eurozone and found significant differences in the adjustment of the bank interest rates following positive and negative changes of the market and central bank rates.
Another line in the relevant literature examined the effect of a monetary regime change in the IRPT process. For example, many studies attempted to evaluate the impact of the introduction of the euro on the degree of pass-through homogeneity in the euro area. Toolsema, Sturm, and de Haan (2002), after introducing structural breaks, showed that although major differences in euro area pass-through exist, there was at least some evidence for convergence of monetary policy transmission. Sander and Kleimeier (2004) found that structural breaks in the pass-through processes of the Euro area members occurred well before EMU and reported increases in size and speed of the pass-through during the latest period examined. However, they found overall heterogeneity in the pass-through mechanism, partially attributing this finding to credit rationing. Marotta (2009), after endogenously identifying structural breaks in the EU pass-through, obtained a pattern of dates suggesting that national banking systems adjust slowly to the new monetary regime; on this basis, he suggested caution in associating structural changes to the introduction of the Euro. De Bondt (2005) examined the pass-through mechanism for the whole Euro area and reported an incomplete pass-though, especially for the short-run. However, he also provided evidence for a quicker bank interest rate pass-through process in the euro area after the introduction of the Euro.
Kok and Werner (2006) applied unit root and cointegration testing in a panel framework on harmonized interest rate data provided by the ECB. They confirmed considerable heterogeneity both of the long-run as well as the short-run adjustment coefficients of the pass-through processes. They explained the heterogeneity of the short-run dynamics on the basis of lack of competition between banks and other structural and cyclical determinants. Bernhofer and van Treeck (2013) analyzed the bank interest rate pass-through in the euro area for the period 1999-2009 by applying the pooled mean group estimator (PMGE). After constraining the long-run pass-through to be homogeneous across countries and allowing country-specific interest rate pass-through in the short-run, they found significant heterogeneity in the short-run pass-through. They also showed that the degree of heterogeneity of the interest rate pass-through has not been improved in the second half of the period examined, suggesting that no significant improvement has been observed under the EMU regime.
Illes and Lombardi (2013) examined the impact of the post-2007 financial crisis on the relationship between monetary policy rates and bank lending rates in the Euro area. They found that a less effective pass-through mechanism was observed after the beginning of the crisis, when near zero policy rates did not lead to lower bank interest rates in some countries. According to them, this result was related to the higher risk premium required by financial intermediaries in financially troubled countries. Blot and Labondance (2013) also examined the impact of the financial crisis on the IRPT in the eurozone using a SUR-ECM model. They showed that the post-2008 financial turmoil has drastically affected the pass-through in the eurozone making it less complete and more heterogeneous across the eurozone members.
Moreover, a number of recent studies also highlight the role of the Eurozone sovereign debt crisis, by showing the effects of diverging sovereign bond yields and country specific structural problems of banks on the divergence of the banking market interest rates and the increased segmentation (Neri, 2013; Acharya & Steffen, 2013; Gennaioli, Martin, & Rossi, 2014; Becker & Ivashina, 2014a, b). On the other hand, the study of von Borstel, Eickmeier, and Krippner (2015) showed that the transmission of conventional monetary policy to bank lending rates did not change with the crisis.
The present study attempts to evaluate the degree of homogeneity of the interest rate pass-through (IRPT) process within the Euro area by focusing both in the short-run adjustment process and the long-run equilibrium. In addition, it aims at assessing the impact of the post 2008 financial turbulence to the euro area IRPT process. The empirical model used here incorporates various structural variables, in an attempt to explain possible heterogeneity in the IRPT process. Another contribution of the study is that it uses the ECB's recently introduced data on the cost of borrowing indicators which are more appropriate for cross-country comparisons relative to banking market interest rates.
Sample and Data Collection
The authors used a panel of recently published monthly data by the European Central Bank (ECB, 2013) on the cost of borrowing referring to a) loans to non-financial institutions b) loans to households for house purchase, c) total long-term lending and d) total short-term lending for the 2003-2013 period. The sample includes all the euro area countries with the exception of Cyprus, Malta, Slovakia and Slovenia, due to lack of data.
Till present, research on the Euro area pass-through process has made quite limited utilization of these recently published indicators. ECB (2013) pointed out that the use of interest rates of specific banking market segments in cross-country analysis suffers from the fact that reported interest rates might show a large degree of cross-country heterogeneity. This is because of the existence of heterogeneity in bank lending practices and banking market products or differences in the data collection and aggregation methodologies across countries. For example, short-term interest rates on business loans reported by the ECB do not take into account lending through overdrafts which is quite popular in some countries. According to ECB (2013), the composite cost of borrowing indicators provided a better picture of the cost of bank lending in each country, enhancing cross-country comparability. However, the ECB only published four aggregate cost of borrowing indicators (as described above) thus limiting the degree of specialization in the analysis.
The majority of the interest rate pass-through studies examined the impact of money market rates on banking market interest rates. The choice of the specific money market rate is usually based on the maturity matching with the banking market rate under examination. De Bondt (2002) chose the specific money market rate that exhibits the highest correlation with the banking market rate in question. Mojon (2000) used a combination of money market rates and 10-year government bond yields.
The majority of the relevant literature evaluates the pass-through of money market rates on bank interest rates without using additional structural determinants. Here, the authors use market interest rates as well as the additional demand and supply side factors affecting the cost of borrowing.
The first step is to test for the existence of unit roots in all variables involved, using the Im, Pesaran, and Shin (2003) panel unit root test (IPS). This test assumes individual error correction terms across countries and has as the null hypothesis that all panels contain unit roots against the alternative that at least one panel is stationary. Acceptance of the null hypothesis makes the use of an error correction methodology in the estimation procedure valid. The IPS test however is too restrictive, as it rejects the unit root hypothesis even when few cases of stationarity are present. For this reason, .results of unit root rejection should be taken with caution. In addition, in order to investigate the existence of long-run relationship, the bounds testing procedure described by Pesaran, Shin, and Smith (2001) is applied in all variables (including those with mixed stationary and non-stationary components). Thus, aiming at detecting long-run relations between variables that might appear to be both stationary and non-stationary.
The second step is to apply the pooled mean group estimator (PMGE) (Pesaran, Shin, & Smith, 1999) which utilizes the error correction methodology after reparametrizing the autoregressive distributed lag form of the equations. Thus, it allows for the estimation of the short-run and possible long-run effects separately in the case of non-stationary regressors while it produces consistent results even in the case of stationarity among the variables. Here, the use of the PMGE allows country specific interest rate pass-through in the short-run while assuming a homogeneous long-run pass-through across countries.
Thus, its key advantage is that it provides specific information concerning the short-run pass-through processes for each country, provided that the restriction of the common long-run pass-through is valid. By restricting the long-run coefficients to be common, there are more available degrees of freedom to estimate the short-run dynamics. However, in case the utilized test for the validity of this restriction (Hausman test) rejects the null hypothesis, the simpler mean group estimator (MGE) is chosen, which allows for different pass-through coefficients both for the short and the long-run. In this case, we report the simple mean of the individual coefficients instead of obtaining a unique coefficient for the whole panel.
Following Pesaran, Shin, and Smith (1999), the initial model is an Autoregressive Distributed Lag of the form ARDL(2,2,2,1,1,1,1), where the cost of borrowing (dependent variable) is regressed on the current and lagged levels of the rest of the explanatory variables, namely the market rate, the spread of the 10 year government bond yields taken as a measure of relative country risk, unemployment rates representing cyclical effects, and banking market supply side variables (such bank liquidity and capital adequacy ratios). The ARDL (2,2,2,1,1,1,1) form of our model is described in equation (1)
[mathematical expression not reproducible] (1)
Were, [R.sub.it] is the relevant cost of borrowing indicator of country i at time t, [M.sub.it] is the euro area market interest rate common to all countries for time t, Sit denotes the spread of the 10-year sovereign bond yields calculated as the difference of each countrys bond yield relative to the German bond yield, ADit is the assets to deposits ratio, C PA it is the capital to assets ratio, CAit is the cash to assets ratio and finally Uit denotes the unemployment rates for each country.
Non-stationarity of the variables requires an estimation procedure that incorporates an error correction mechanism. This way, the long-run equilibrium is distinguished from the short-run dynamics and the initial ARDL model can be reparametrized as in equation 2.
[mathematical expression not reproducible] (2)
[mathematical expression not reproducible]
In equation 2, AX denotes the first difference of each corresponding explanatory variable X (i.e. [DELTA]X=[X.sub.t] -[X.sub.t-1]), [phi] is the error adjustment coefficient, while [phi], x, v, p., v, [zeta] o, [pi], [rho], [sigma] and [[theta].sub.1]... [[theta].sub.6], are parameters to be estimated.
As aforementioned, when the PMGE is applied, the long-run coefficients are assumed to be common across countries. In other words, the PMGE is derived from the restricted version of the equation (2) where the coefficients [[theta].sub.1i]...[[theta].sub.6i] are common across countries.
Here, the aim is to evaluate the impact of the post-2008 financial crisis on the interest rate pass-through mechanism. For this purpose, the authors derived separate PMG and MG estimators for two sub-samples, i.e. for the period before the outbreak of the financial crisis and the period after. Since the exact time of the initiation of the financial crisis impact might be different for various countries, we look for the presence of structural breaks in the model for each country separately. For this purpose, we use the Quandt likelihood ratio test (QLR) (Quandt, 1960) to derive the (main) structural break endogenously and we use the critical values for a chi square distribution provided by Stock and Watson (2003). After the derivation of the exact period of structural break for each country separately, we use the mean period as a single date for splitting the sample into the pre-crisis and post-crisis sub-periods.
Since the model employed here uses additional explanatory variables next to market interest rates (unlike the majority of the empirical work on the interest rate pass-through), it is highly possible to observe multiple breaks. Thus, an appropriate methodology would be the one that indicates the most significant break, assuming that this will be closely related to the extraordinary event of the 2008 financial crisis.
Following the approximation of the structural break point for the whole panel, the authors proceed with the estimation of the model using both the pooled mean group and the mean group estimators. As described above, the pooled mean group is mostly preferable because it allows more degrees of freedom for the estimation of the individual error adjustment coefficients [phi] and the rest short-run dynamics. A Housman test is used to establish whether the PMG estimator is consistent. Rejecting the null hypothesis of the Housman test suggests that the MGE is more consistent. Moreover, such testing of the common long-run coefficients restriction is an indirect way of testing the homogeneity of the long-run pass-through relation across countries. In the case where the PMG estimator is consistent, one can conclude that the heterogeneity of the pass-through comes mainly from the individual error adjustments and the rest short-run dynamics indicating differences in the levels of competition and other structural differences. Otherwise, heterogeneity across countries can be assumed to originate from both individual long-run relations and short-run dynamics.
Concerning the short-run dynamics, we are mostly interested in the error adjustment coefficients. Both PMG and MG estimators allow for the error adjustment coefficients [phi] to vary across countries. Measuring the standard deviation of the error adjustment coefficients across countries indicates the degree of dispersion of the short-run dynamics of the pass-through mechanism across the regions. The error adjustment measures how quick the response of an individual countrys bank interest rates is to changes in the market inte rest rates. The greater the standard deviation the less homogenous is the individual responses to these changes. One approach could be to search for potential trends on the value of standard deviation applying a rolling procedure for the estimation of the error adjustment coefficients [phi] and to observe the evolution of the standard deviation through time. This study intends to draw conclusions for distinct periods, namely the pre-crisis period and the period after its onset, and report the standard deviations calculated from the two sub-periods resulting from the break.
Here, the authors employ a principal component analysis (as suggested by the ECB, 2013) to construct a representative indicator for the money market rates. For this purpose, we use the EONIA rate and the 1-month, 3-months, 6-months and 1-year Euribor rates. Since the first principal component captures almost 99% of the variation of market rates, we do not consider any additional components for our analysis.
The utilized model also includes additional relevant structural variables that might affect the cost of borrowing from banks. Unemployment rates are used as a measure of aggregate demand conditions in each country; the ten-year government bond yield spread of each country vis-a-vis Germany is used to capture country risk which, in turn, relates to default risk in banking; finally, supply side conditions in the banking market are captured with the use of cash to asset ratio, capital to asset ratio and asset to deposit ratio (measuring the degree of liquidity, leverage and asset growth respectively). All ratios used are calculated with data published by the ECB.
The tests, as well as the analysis, programming and the estimation of the models was done using the statistical software package STATA. Estimation using the PMG and MG was done following Blackburne and Frank (2007).
Table 1 presents the results of an Im, Pesaran & Shin test (IPS test) used here to test for the existence of unit roots in the whole panel. IPS test with the inclusion of lags, is appropriate when the time dimension T tends to infinity before the cross section dimension N (i.e. the number of countries). By using this test, the error correction adjustment coefficient is allowed to vary across panels. As shown in Table 1, the hypothesis of a unit root cannot be rejected for all the variables except for the long-term cost of borrowing and the cash to assets ratio. It should be noted here, that the alternative hypothesis of this test, is that at least one panel is stationary. Thus, the test is rather strict towards rejecting the unit root hypothesis. According to Pesaran, Shin, and Smith (1999), the procedure of the pooled mean group estimation is consistent under the assumption that all model regressors are either stationary or non-stationary. In our case, all regressors are non-stationary except for the variable bank cash to assets ratio. However, as it is shown below in model estimation results, the cash to assets ratio variable does not have any significant effects in the long-run relationships.
Concerning the bounds testing procedure, after pooling the data and arranging them in order to apply the test appropriately, we first search for the number of lags that minimize the Schwarz criterion. After determining that the criterion is minimized using just one lag in our model, we next apply the bounds test, calculating the F statistic on the hypothesis that all coefficients of the variables in the long-run relation are jointly zero. The F statistic is 4.17 and the variables in the long-run relation are m=7. Using this information and referring to the critical values from the tables of Pesaran, Shin & Smith (2001) for no trend and unrestricted constant, we see that the value of our statistic for k=6 (k+1=m) is greater than the upper bound. This leads us to conclude that there is a significant long-run relationship between the variables across the whole panel.
Next, the authors use the Quandt Likelihood Ratio test, or QLR test (Quandt, 1960) which is a rolling Chow test, also known as SupF test to search for possible breaks in the long-run relationship (eq. 2) for each country and for all cost of borrowing indicators. The test results are shown in table 2. For each country the QLR statistic is the maximum of the Chow test statistics produced after each date is tested as a possible break date. In our case the QLR statistics appear significant in all cases at the 99% confidence level (using the critical values from Stock and Watson (2003)) indicating the existence of structural breaks. It needs to be stressed that by applying the QLR for the whole sample period, a single break (the one with the highest significance) is identified. Since our model uses multiple regressors, the probability of existence of multiple breaks increases. However, since the post-2008 financial crisis has by far been the most serious financial disturbance for all countries for the period under examination, we expect that the most significant structural break would be related to this event.
The aforementioned QLR test procedure, not only provides information on the existence of statistically significant structural change in coefficients, but also points out the date at which the break took place for each individual country. Then we approximate a single date for the structural break for the whole sample by taking the average period of the dates of individual country breaks, as shown in Table 2. It is evident that the key structural breaks for the vast majority of the cases lie between 2008 and 2010 and are thus closely related to the sovereign debt crisis that unfolded during that period. Thus, it is possible to use the average structural break dates in splitting the whole sample in pre-crisis and post-crisis sub-periods and estimate the model for both of them.
Tables 3 and 4 show the estimation results of the mean group and pooled mean group estimators for the pre-crisis and post-crisis periods respectively. In each table, the estimated long-run equilibrium relationships are reported as well as the error adjustment coefficient [phi] for all cost of borrowing indicators. Since the main interest of this study is the long run relation and the error adjustment mechanism, the estimated coefficients of the first differences of variables and their lags are not reported. However, the inclusion of the first differences of variables and their lags is necessary for estimating the models consistently. The lag order selection of the first differenced variables was based on their significance levels, as increasing the lag order showed reduced significance. Moreover, increasing the lag order, would consume degrees of freedom, as some or all of the models' parameters are estimated for each country individually. Finally, the lag structure used is justified by the fact that the residuals showed no significant autocorrelation.
In each table, the chi square statistic of the Hausman test and its significance level is reported at the bottom. The Hausman test examines whether the restriction of the common long-run equilibrium coefficient across countries is valid. A rejection of the common long-run coefficients of the regressors across countries calls for the use of mean group estimation. The mean group estimator allows not only the error adjustment and the short-run coefficients to vary across countries, but also allows for different long-run coefficients. Following the estimation of individual equations, a simple average of the coefficients across countries is reported.
Thus, while in the case of validity of the pooled mean group estimator the main source of cross-country monetary policy transmission heterogeneity is the error adjustment mechanism and the other short-run effects, in the case of validity of mean group estimator the source comes both from the long-run relationships and the short-run dynamics. In Tables 3 and 4, in the case where the p value of the Hausman test is over 0.05, the null of no systematic differences between the coefficients of the two estimators is accepted, thus, the pooled mean group estimator is preferable. When, on the other hand, the test rejects the null, the mean group estimator is preferred as more consistent.
The Hausman test shown in Table 3 suggests that for the pre-crisis period the mean group estimator is the preferred estimator for all but one cost of borrowing indicators, the exception being the case of loans to non-financial firms (NFIs). This means that during the pre-crisis period the cross-country heterogeneity of the impact of the explanatory variables on the cost of borrowing stem from both the long-run as well as the error adjustment and the other short-run effects. In the case of the NFIs cost of borrowing cross-country, only, heterogeneity in the pass-through process stems mainly from the short-run effects, suggesting a homogeneous long-run process.
The Hausman test statistic for systematic differences in coefficients of PMG and MG and the p-values appear at the bottom line. Bold letter coefficients refer to the significant ones of the consistent estimation method suggested by the Hausman test. The Hausman specification test, has as null hypothesis that the coefficients of MG and PMG are consistent, but PMG is efficient. The alternative hypothesis is that the coefficient of MG is consistent, while the one of PMG is not. Hausman test can give a negative value, when the variance of the parameter estimate of PMG is greater than that of MG. Here, the consistency of PMG is tested, conditional on the fact that MG is not just consistent but also more efficient (as it has smaller variance and hence negative test statistic). It turns out that, a negative Hausman test indicates that MG is consistent.
According to Marotta (2009), the cross-country heterogeneity of the short-run adjustment coefficients in the pass-through process could be partially attributed to the low level of cross-border competition among banks. The error adjustment coefficients [phi] measures the speed at which banks adapt as a response to changes in bank market conditions. Even under diverse market conditions across countries, bank competition might increase the speed at which banks respond to bank market changes. On the other hand, cross-country heterogeneity due to long-run equilibrium relationships suggests the existence of significant heterogeneity in the degree of completeness of the pass-through process.
The pass-through is considered as complete when the coefficient of the money market rates variable equals to 1. As Table 3 shows, the size of the estimated coefficients for the money market variable (M) is quite low (less than 1) indicating incomplete pass-through to bank interest rates. Moreover, the rejection of the hypothesis of homogeneous long-run coefficients across countries in three out of four loan categories indicates statistically significant differences in the completeness of the pass-through and the equilibrium relations across countries during the pre-crisis sub-period.
Table 4 shows the estimation results for the post-crisis sub-period. In this case, contrary to the results for the pre-crisis period, the common long-run coefficients restriction cannot be rejected in three out of four loan categories with the household loans being the only exception. This leads to the conclusion that during the post-crisis period cross-country heterogeneity in the long-run impact of the explanatory variables has been limited. Therefore, it can be claimed that during that period the degree of completeness of the pass-through is more homogeneous across countries and that cross-country differences in the pass-through processes mainly stem from the short-run effects and other differences in the individual factors affecting the bank rates. In this case, convergent financial conditions across countries would result into convergent cost of borrowing in the long-run, no matter what differences in the speed of adjustment may exist.
The Hausman test statistic for systematic differences in coefficients of PMG and MG and the p-values appear in the bottom line. Bold letter coefficients refer to the those that are significant based on the consistent estimation method suggested by the Hausman test as presented in Table 3.
The main interest, is in the estimation results of the consistent estimation method (either mean group or pooled mean group) as suggested by the Hausman tests. First, Tables 3 and 4 suggest that the error correction adjustment coefficient is significant in all cases for both sub-periods and for both estimators. Therefore, the existence of a long-run equilibrium between the bank interest rates and the market rates and the rest of the explanatory variables cannot be rejected in all cases. However, the long-run relationships estimated for the two sub-periods show notable differences:
a) For the pre-crisis sub-period, the mean group estimation for the cost of borrowing of loans to households, long-term loans and short-term loans reveals that the structural variables (i.e. assets to deposits, capital to assets, cash to assets and unemployment rates) are not significant determinants of bank interest rates. On the other hand, we can see that the market rates are significant in all cases, while the sovereign bond spreads are only significant in the case of the long-run cost of borrowing indicator. In the case of loans to NFIs, the pooled mean group estimation shows that all variables are significant with the exception of capital to assets and cash to assets ratios. In a nutshell, the estimation results for the pre-crisis sub-period indicate that the pass-through process was close to the description of the traditional pass-through models which relate interest rate changes merely to changes in policy rates or money market rates through the market rates.
b) For the post-crisis sub-period, as mentioned earlier, the pooled mean group estimation is preferred, with the exception of the case of loans to households. In the cost of borrowing categories where the pooled mean group estimator was applied, the structural variables used are significant determinants of the variation of the interest rates across countries. Therefore, the traditional pass-through model is not adequate to capture all factors affecting the pass-through process across countries. Specifically, besides money market rates, sovereign bond spreads, unemployment rates and the balance sheet ratios are significant factors explaining cross-country variability of interest rates in the majority of the cases examined. The picture however is different for the loans to households, where the mean group estimator yields significant results only for the money market rate and the sovereign spreads variables.
As expected, the coefficients of the market rate variable in all the cases examined (for both sub-periods) are positive but well below 1, indicating less than complete pass-through. The coefficient of the sovereign spreads variable is positive in all cases with the exception of loans to NFI s for the pre -crisis period. Therefore it can be claimed that bank interest rates are positively related to sovereign spreads, the latter being a measure for country risk.
Similarly, unemployment rates, when significant, relate positively to bank interest rates. Higher unemployment signifies deepening recession, deterioration of expected income by firms and households and thus widening risk premia in bank lending. The coefficient of the assets to deposits ratio is negative in all but one case: a rising assets to deposits ratio might relate to loose supply conditions in bank lending and thus to falling cost of borrowing. Moreover, capital to assets ratio is found to be positively related to the cost of borrowing. As capital to assets ratio increases, the degree of leverage in banking activities falls putting pressure on banks towards increasing their return to assets (and thus lending rates) in order to avoid worsening return to equity ratios. Finally, the variable of cash to assets ratio as a measure of liquidity was not found significant in almost all cases examined.
Next the authors examined the impact of the financial crisis on the short-run adjustment of the interest rate pass-through process across countries. For this purpose they report the crosscountry standard deviations of the error adjustment coefficients [phi] for the two sub-periods in Table 5.
The overall conclusion is that there has been an increase in the dispersion of the error adjustment coefficient [phi] in the post-crisis period, especially in the cases of the cost of borrowing for total short-term and total long-term lending. In other words, the speed at which countries' cost of borrowing respond to changes in the market rates varies considerably more during the post-crisis period. On the other hand, this conclusion does not hold for the cost of borrowing of loans to NFIs.
This study examines the interest rate pass-through process in the euro area banking market and the impact of the post-2008 financial turmoil on this process using the recently produced cost of borrowing indicators by the ECB (2013). In order to investigate the factors behind the divergence in the interest rate pass-through process, we used a typical interest rate pass-through model augmented with key structural variables which capture demand and supply conditions in banking markets across counties.
The findings suggest that there has been an increase in the heterogeneity of the speed of the short-run adjustment mechanism of the pass-through process across countries during the post-crisis period. Increased cross-country heterogeneity is also observed in the long-run coefficients of the money market rates pass-through to cost of borrowing in bank lending during the same period. In other words, it seems that the financial crisis has increased the cross-country heterogeneity in the degree of completeness of the interest rate pass-through process.
However, the error correction adjustment coefficient in all the cost of borrowing categories examined appears negative in the post-crisis period. This suggests that a short-run adjustment process towards interest rate pass-through convergence in the long-run is at work. It is also a positive finding for the convergence process that the statistical analysis used here does not reject the hypothesis of the common long-run coefficients of the pass-through process across euro area countries. This result suggests that a rather homogenous pass-through process prevails in the long-run.
Another important finding of the present study is that the structural variables used here to capture demand and supply conditions of bank loan markets have significant impact on the interest rate pass-through process especially for the post crisis period. Thus, part of the increased heterogeneity of the pass-through process across countries after the outbreak of the financial crisis can be attributed to divergent demand and supply conditions prevailing during this period. Therefore, it can be supported that the traditional pass-through models usually adopted by the empirical literature, in which the variability of lending rates is merely attributed to money market rate variability, cannot give the full picture of the interest rate pass-through process within the euro area.
Overall, the key conclusion emanating from the study is that the interest rate convergence process within the euro area could be strongly enhanced with the convergence of banking market conditions among the euro area members. The latter calls for the adoption of policies that aim to provide a viable resolution to the euro area debt crisis and to further improve the degree of banking market integration within the euro area.
Acharya, V., & Steffen, S. (2013). The Greatest Carry Trade Ever? Understanding Eurozone Bank Risks, CEPR Discussion Papers, 9432.
Barigozzi, M., Conti, A. M., & Luciani, M. (2013). Do euro area countries respond asymmetrically to the common monetary policy? Temi di discussione (Economic working papers) 923, Bank of Italy, Economic Research and International Relations Area.
Becker, B., & Ivashina, V. (2014). Cyclicality of Credit Supply: Firm Level Evidence, Journal of Monetary Economics, 62, 76-93.
Becker, B., & Ivashina, V. (2014). "Financial Repression in the European Sovereign Debt Crisis", Swedish House of Finance, Research Paper No. 14-13, Paris December 2014 Finance Meeting EUROFIDAI - AFFI Paper.
Bernhofer, D., & van Treeck, T. (2013) New evidence of heterogeneous bank interest rate pass-through in the euro area. Economic Modelling, 35 418-429.
Blackburne, E. F. III, & Frank, M. W. (2007). Estimation of nonstationary heter?geneous panels. The Stata Journal, 7(2) 197-208.
Blot, C., & Labondance, F. (2013). Business lending rate pass-through in the Eurozone: monetary policy transmission before and after the financial crash. Economics Bulletin 33(2), 973-985.
Cottarelli, C., & Kourelis, A. (1994). Financial structure, bank lending rates, and the transmission mechanism of monetary policy. IMF Staff Papers, 414 587-623.
De Bondt, G. (2002). Retail bank interest rate pass-through: new evidence at the Euro area level. ECB Working Paper Series, 0136.
De Bondt, G. (2005). Interest rate pass-through: empirical results for the Euro Area. German Economic Review, 6(1) 37-78.
European Central Bank (2013). Assessing the retail bank interest rate pass-through in the Euro area at times of financial fragmentation. ECB Monthly Bulletin, 08.
Gennaioli, N., Martin, A., & Rossi, S. (2014). Sovereign Default, Domestic Banks, and Financial Institutions. Journal of Finance, 69 (2), 819-866.
Hausman, J. A. (1978). Specification Tests in Econometrics. Econometrica. 46 (6): 1251-1271.
Heinemann, F., & Schuler, M. (2002). Integration benefits on EU retail credit markets--Evidence from interest rate pass-through. ZEW Discussion Papers 02-26.
Illes, A., & Lombardi, M. J. (2013). Interest rate pass-through since the financial crisis. BIS Quarterly Review, 09.
Im, K. S., M., Pesaran, H., & Shin, Y. (2003). Testing for unit roots in heterogeneous panels. Journal of Econometrics, 115, 53-74.
Karagiannis, St., Panagopoulos, Y. , & Vlamis, Pr. (2010). Interest rate pass-through in Europe and the US: Monetary policy after the financial crisis. Journal of Policy Modeling 32(3), 323-338.
Kok, C., & Werner, T. (2006). Bank interest rate pass-through in the euro area: a cross country comparison. ECB Working Paper Series, 0580.
Marotta, G. (2009). Structural breaks in the lending interest rate pass-through and the euro. Economic Modelling, 26(1), 191-205.
Mojon, B. (2000). Financial structure and the interest rate channel of ECB monetary policy. ECB Working Paper Series, 0040.
Neri, S. (2013). The impact of the sovereign debt crisis on bank lending rates in the euro area, Questioni di Economia e Finanza (Occasional Papers) 170, Bank of Italy, Economic Research and International Relations Area.
Pesaran, M., Shin Y., & Smith R. P. (1999). Pooled mean group estimation of dynamic heterogeneous panels. Journal of the American Statistical Association, 94(446), 621-634.
Pesaran, M., Shin, Y. , & Smith, R. (2001). Bounds Testing Approaches to the Analysis of Level Relationships. Journal of Applied Econometrics, 16(3), 289-326.
Quandt, R.E., (1960).Tests of Hypotheses that a Linear System Obeys Two Separate Regimes. Journal of the American Statistical Association, 55, 324-330.
Sander H., & Kleimeier, St. (2001). Asymmetric Adjustment of Commercial Bank Interest Rates in the Euro Area: An Empirical Investigation into Interest Rate Pass-Through. Kredit und Kapital, 2(1-32), 161-192.
Sander, H., & Kleimeier, S. (2004). Convergence in Eurozone Retail Banking? What Interest Rate Pass-Through Tells Us about Monetary Policy Transmission, Competition and Integration. Journal of International Money and Finance, 23(3), 461-492.
Stock, J. H., &Watson, M.W. (2003). Introduction to Econometrics. Boston: Addison Wesley Longman
Toolsema, L. A., Sturm, J. E., & de Haan, J. (2002). Convergence of pass-through from money market to lending rates in EMU countries: new evidence. CCSO Working Papers 200206.
Von Borstel, J., & Eickmeier, S., & Krippner, L. (2015). The interest rate pass-through in the euro area during the sovereign debt crisis. Discussion Papers, no 10/2015, Deutsche Bundesbank Research Centre.
University of Macedonia, Greece
About the Authors:
George Michalopoulos is Associate Professor at the Department of Accounting and Finance, University of Macedonia, Thessaloniki, Greece. His published research work is in the area of International Banking, International Monetary Relations and European Financial Integration.
Konstantinos Tsermenidis is a PhD holder from the Department of Accounting and Finance, University of Macedonia, Thessaloniki, Greece. His published research work and main academic interests lie in the fields of European Banking Integration, Economic Convergence and Econometric Analysis.
Table 1 Im-Pesaran-Shin Test for Unit Root for the Whole Panel Wt-bar P Cost of borrowing - NFI -1.13 0.13 Cost of borrowing - Households -0.76 0.22 Cost of borrowing - Long-term Loans -2.12 0.02 Cost of borrowing - Short-term Loans -1.34 0.09 Money Market Rate 0.96 0.83 Sovereign Spread -0.29 0.39 Assets to Deposits ratio -0.48 0.32 Capital to Assets ratio 5.25 1.00 Cash to Assets ratio -1.73 0.04 Unemployment Rate 1.42 0.92 Notes: Number of lags chosen according to AIC criterion between 0 and 5, null hypothesis: all panels contain unit roots against alternative that at least one panel is stationary. [W.sub.t-bar] statistic is the IPS test statistic when the time dimension (T) tends to infinity faster than the cross section dimension, (i.e. the number of countries N) (Im, Pesaran & Shin (2003). Table 2 QLR Test for Breaks in the Coefficients of All Independent Variables on the Four Categories of the Cost of Borrowing for Each Individual Country NFI loans Household loans QLR Date QLR Date Austria 18.22 2006/04 20.75 2008/03 Belgium 25.42 2005/05 28.40 2007/12 Finland 9.42 2009/06 24.32 2009/04 France 9.58 2010/01 43.21 2010/03 Germany 35.27 2007/04 36.64 2008/11 Greece 26.01 2010/07 39.64 2009/03 Ireland 8.34 2010/11 14.89 2010/03 Italy 20.17 2009/12 37.50 2007/03 Luxemburg 7.58 2011/02 47.24 2009/04 Netherlands 8.94 2009/01 19.82 2009/05 Portugal 24.10 2009/08 33.18 2008/11 Slovenia 199.21 2008/05 14.86 2010/01 Spain 8.44 2011/01 28.98 2009/12 Average Date 2009/02 2009/02 Long-term loans Short-term loans QLR Date QLR Date Austria 15.23 2007/01 18.40 2006/04 Belgium 46.79 2010/02 23.97 2005/05 Finland 12.49 2008/10 16.46 2009/06 France 40.22 2010/03 6.86 2009/04 Germany 36.18 2009/05 29.04 2007/03 Greece 19.30 2009/02 24.07 2010/07 Ireland 3.50 2011/05 19.98 2010/09 Italy 24.70 2007/04 17.10 2009/12 Luxemburg 4.18 2008/09 7.05 2011/11 Netherlands 15.65 2008/08 7.63 2009/02 Portugal 2.34 2009/09 30.63 2009/08 Slovenia 9.87 2010/01 204.07 2008/05 Spain 6.61 2008/10 10.03 2009/12 Average Date 2009/03 2009/01 Notes: The QLR test is the maximum F-statistic from the Chow test over a certain sample period. The date of the maximum break for each individual country is set at the point where the corresponding F statistic is maximum. Average date, is the average break date of all countries. Table 3 Results of the Pooled Mean Group and Mean Group Estimation of Equation (2) for Four Cost of Borrowing Indicators for the Pre-Crisis Period Loans to NFI (Pre-2009/02 period) Household loans (Pre-2009/02 period) PMG Coef. PMG Coef. MG Coef. 9 -0.22 (**) MG Coef. -0.14 (***) -0.32 (***) M 0.61 (***) -0.56 (***) 0.57 (***) 0.39 (***) S -0.23 (**) 0.50 (***) -0.43 (***) 0.20 AD -0.49 (***) -0.26 -0.58 (***) -1.08 CPA 1.77 -0.87 -5.57 (*) 8.63 CA 0.32 1.72 -1.47 7.46 U 0.07 (***) -4.07 0.03 (***) -0.02 Hausman Chi sq. 9.34 P-0.07>chi sq. 0.16 [phi] -0.29 (***) -0.59 (***) -0.22 (***) -0.50 (***) M 0.33 (***) 0.35 (***) 0.62 (***) 0.42 (***) S 0.13 0.61 (**) -0.17 (*) -0.01 AD -0.57 -0.10 -0.39 (***) -1.27 CPA 18.83 (***) 14.48 1.53 -8.96 CA -2.38 -9.71 0.22 -2.52 U -0.26 (***) -0.06 0.08 (***) -0.20 Hausman Chi sq. -5.9 8 P>chi sq. 0.00 Chi sq. 20.6 6 P>chi sq. 0.00 Sources: Data from ECB. Notes: Variables as described in eq. (1); asterisks next to estimated coefficients, indicate the significance level. Asterisks (*), (**), (***) indicate p-values less than 0.05, 0.01 and 0.001, respectively. Table 5 Cross Country Standard Deviations of the Error Adjustment Coefficients [phi] for All Four Cost of Borrowing Indicators for the Pre- and the Post-Crisis Periods. NFI loans Households loans Long-term loans Pre-2009/02 Pre-2009/02 Pre-2009/03 PMG MG PMG MG PMG MG 0.28 0.32 0.11 0.12 0.29 0.29 Post-2009/02 Post-2009/02 Post-2009/03 PMG MG PMG MG PMG MG 0.21 0.32 0.11 0.19 0.40 0.44 NFI loans Short-term loans Pre-2009/02 Pre-2009/01 PMG PMG MG 0.28 0.16 0.25 Post-2009/02 Post-2009/01 PMG PMG MG 0.21 0.27 0.34
|Printer friendly Cite/link Email Feedback|
|Author:||Michalopoulos, George; Tsermenidis, Konstantinos|
|Publication:||International Journal of Business and Economics Perspectives (IJBEP)|
|Date:||Sep 22, 2019|
|Previous Article:||PRODUCTIVITY IN SERVICE, MANUFACTURING, AND HYBRID INDUSTRIES: AN EMPIRICAL STUDY.|
|Next Article:||PREDICTING BANK FAILURES: THE NECCESITY OF BANK CAPITAL REQUIREMENTS.|