Asymmetric Adjustment of Commercial Bank Interest Rates in the Euro Area: An Empirical Investigation into Interest Rate Pass-Through

Sander, Harald; Kleimeier, Stefanie Article Asymmetric Adjustment of Commercial Bank Interest Rates in the Euro Area: An Empirical Investigation into Interest Rate Pass-Through Kredit und Kapital Provided in Cooperation with: Duncker & Humblot, Berlin Suggested Citation: Sander, Harald; Kleimeier, Stefanie (2002) : Asymmetric Adjustment of Commercial Bank Interest Rates in the Euro Area: An Empirical Investigation into Interest Rate Pass-Through, Kredit und Kapital, ISSN 0023-4591, Duncker & Humblot, Berlin, Vol. 35, Iss. 2, pp. 161-192, https://doi.org/10.3790/ccm.35.2.161 This Version is available at: https://hdl.handle.net/10419/293457 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Kredit und Kapital, Heft 2/2002 Seiten 161 - 192 Asymmetric Adjustment of Commercial Bank Interest Rates in the Euro Area: An Empirical Investigation into Interest Rate Pass-Through By Harald Sander, Cologne, and Stefanie Kleimeier, Maastricht* Shooting at a moving target in the fog is no easy task. Dornbusch, Favero and Giavazzi (1998) The markets are asymmetric; we are not. Alan Greenspan I. Introduction Since January 1, 1999 the new European Central Bank (ECB) has to conduct a "one-size-f its-all" monetary policy based on her assessment of the average economic conditions of the member countries of the European Monetary Union (EMU). Next to the usual issues and controversies in monetary policy making this implies three new challenges: (1) determining the appropriate average monetary policy in case of diverging economic conditions in the euro area, (2) dealing with possible asymmetric effects of that monetary policy in different member countries, i.e. a divergent monetary transmission mechanism which (3) is most likely subject to dramatic changes (convergence?) as financial market integration and restructuring alongside EMU evolves. While the first challenge has always been at the heart of the controversies about a common currency, the second issue has only recently become an important topic in empirical research. While the latter development is to be welcomed, the third challenge should remind us that judgements about the workings of the monetary mechanism that are based on past data could be misleading in the * The authors are thankful for valuable comments from an anonymous referee. They also wish to thank the discussants and participants of the 3rd Conference of the Swiss Society For Financial Market Research in Zurich, April 7, 2000 and of the "Financial Structure, Bank Behaviour and Monetary Policy in the EMU" conference held at Groningen University, October 5-6, 2000 for helpful comments on an earlier version of this study. As usual, all remaining errors are those of the authors. S. Kleimeier would like to acknowledge the financial support from METEOR. Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 162 Harald Sander and Stefanie Kleimeier context of a regime change. However, this does not mean that analysing past data is of no use. Our study, therefore, focuses on the latter two challenges by providing new evidence on the financial market side of the monetary mechanism in European Union (EU) countries. In particular, we are examining the pass-through of money market rates to commercial bank lending rates by allowing for the existence of and possible changes in a long-run equilibrium relationship between these two interest rates in the period preceding EMU. We extend on the existing literature by considering a number of different symmetric and asymmetric adjustment mechanisms within and across countries to gain a deeper understanding of the nature and diversity of the financial sector in EU countries and its implication for monetary policy making in the EMU. In the early 1990, arguably following up on two publications by the Bank for International Settlements (BIS 1994, 1995), a number of studies have investigated asymmetric responses of output to monetary policy innovations across countries which may complicate the implementation of a single monetary policy in the euro area (e.g. Britton and Whitley 1997, Ramaswamy and Sloek 1997, Barran, Coudert and Mojon 1997, Dornbusch, Favero and Giavazzi 1998). While most studies argue the case of asymmetric effects across countries, it is not undisputed that the evidence provided so far is clearly in favour of this hypothesis. For example, Kieler and Saarenheimo (1998) argue that the "econometric evidence does not provide a coherent picture of such differences". They attribute the failure to provide clear econometric evidence to the issues of correctly identifying monetary policy actions1 and their causal effect on the economy in the "current (or more precisely, historical) set-up". This way the authors provide evidence in favour of no statistically significant differences in monetary transmission for Germany, France and the United Kingdom. However, most empirical papers are based on an estimation period from the 1970s up to date, a time period over which one has to account for changing and differing exchange rate regimes, the 1992/93 EMS crises, and a number of exchange rate re-alignments that all have had an impact on the workings of monetary policy. While it is very clear that these differences will disappear with the adoption of a single currency2, it is less clear to what extent there will be a convergence in the monetary transmission mechanism itself. i Kieler and Saarenheimo (1998) use a one percentage point increase in the three month money market rate which is sustained over a period of four years, or, more precisely, they simulate a series of monetary shocks such that they produce a sustained one percentage point increase in the money market rate. Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 Asymmetric Adjustment of Commercial Bank Interest Rates in the Euro Area 163 Since Franco Modigliani (1963) the monetary mechanism has been described to consist of two parts: the financial market reaction and the wage-price mechanism. While the studies reviewed above have been examining the impact of monetary policy on the real economy, our study concentrates on the financial market reaction. There are a number of good reasons to do so: First, given the high proportion of bank finance in Europe relative to the USA and UK as shown in Table 1, the "lending channel" is an important element in the monetary mechanism in Europe (e.g. Bernanke and Gertler 1995, Kashyap and Stein 1993). If loans and bonds are imperfect substitutes in the balance sheets of banks and firms, and firms cannot simply access the capital markets but have to rely on bank finance, the transmission of monetary policy impulses is necessarily linked to bank behaviour. Second, if the structure of the financial system matters as a "conveyer" of monetary policy, these structural differences can lead to asymmetries in European banking market reaction and thus monetary policy transmission. In Germany, for example, the close bank-firm relationship tends to weaken the money market rate - lending rate link, while in economies like the British the like is known to be much more direct. Third, while there is evidence that the wage-price process is different across Europe, the Lucas principle suggests that this very process may adopt to the European focus of the ECB's monetary policy (Dornbusch, Favero and Giavazzi 1998). Banking markets, however, may be more resistant to convergence. E.g. Cecchetti (1999) argues that "differences in financial structure are the proximate cause for these national asymmetries in the monetary policy transmission mechanism" and adds that "unless legal structures are harmonised across Europe, financial structures will remain diverse, and so will the monetary transmission mechanism". In a similar vein in a recent ECB working paper Mojon (2000) argues in favour of concentrating his analyses on the passthrough of interest rate innovations to retail banking rates: "National segmentation in the European retail banking industry may remain significant in spite of EMU, because retail banking involves heavy investments in brand names, in a network of branches and in relationships with customers (Gual 1999) as well as country-specific legal expertise (Cecchetti 1999). As a consequence, the pass-through from policy-controlled interest rates to bank retail interest rates and the effects of those rates on spending decisions may remain country specific. This potential source of asymmetry across countries is particularly relevant in the euro area where bank rates are a key determinant of the cost of capital and the yield on savings." 2 The studies by Dornbusch, Favero and Giavazzi (1998) and Kieler and Saarenheimo (1998), respectively, explicitly account for intra-EMU exchange rate effects. Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 164 Harald Sander and Stefanie Kleimeier Table 1: The Relative Importance of Bank Finance in Europe Panel A: Country Specific Characteristics in 1996 Country Market Capitaliza-Corporate Debt Bank Loans as tion as a Percentage as a Percentage a Percentage of all of GDP of GDP Forms of Finance EU Member Countries Austria 15 46 65 Belgium 45 60 49 Denmark 41 105 25 Finland 50 34 39 France 38 49 49 Germany 29 58 55 Greece 20 3 48 Ireland 18 13 80 Italy 21 37 50 Netherlands 96 48 53 Portugal 23 19 62 Spain 42 11 58 Sweden 99 73 32 United Kingdom 150 45 37 Other Countries Japan 67 39 59 USA 111 64 21 Panel B: Euro-Area Characteristics in June 1999a Euro Area USA Japan Bank Loans 100.4 48.4 107.0 Outstanding domestic debt securities 88.8 164.6 126.5 - issues by corporates 3.3 29.0 14.6 - issued by financial institutions 31.0 45.4 18.8 - issued by the public sector 54.5 90.2 93.1 Stock Market 71.1 163.3 137.7 Capitalization Source: Cecchetti (1999) for Panel A, ECB Monthly Bulletin, January 2000, for Panel B. a All data are in percent of GDP and are given for June 1999 except for stock market capitalization which are for October 1999. Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 Asymmetric Adjustment of Commercial Bank Interest Rates in the Euro Area 165 Until recently, the literature has often neglected the explicit empirical analysis of the pass-through of monetary policy in the financial sector. Exceptions are e.g. Cottarelli and Kourelis (1994) who focus on the impact of money market rate and policy rate changes on the lending rate, Cottarelli, Ferri, and Generale (1995), BIS (1995) and International Monetary Fund (1996), Sander and Kleimeier (2000), and Toolsema et al. (2000, 2001). Dornbusch et al. (1998) review the pre-1998 literature with respect to the financial market reaction in potential EMU member countries and find that the characteristics of the financial system "go some way towards explaining the observed asymmetries in the transmission mechanism". Our study extends on this literature in four important aspects: 1. The traditional pass-through model that utilizes an autoregressive distributed lag specification is extended by an error correction mechanism, which drives the rates back toward their long-run equilibrium relationship. 2. This long-run equilibrium relationship is analysed in a cointegration approach that tests and allows for structural breaks in order to examine the impact of changing conditions on financial market performance so far. 3. Recent research has shifted toward analysing asymmetric adjustment in interest rates (see Tong 1983, Scholnick 1996 and 1999, Balke and Fomby 1997, Enders and Granger 1998; Baum and Karasulu 1998, Ender and Siklos 2000). We therefore test for cointegration in the presence of asymmetric adjustment of interest rates. 4. After these three steps, we select for all EU countries the best-specified error correction pass-through model and obtain impact multiplier, long-run multiplier, and speed of adjustment coefficients that incorporate the relevant symmetric or asymmetric autoregressive decay. While our findings largely confirm the results of earlier pass-through studies, such as the lack of convergence in the financial part of the monetary transmission mechanism, we provide these results within a more refined empirical analysis, which allows us to also identify the nature of the adjustment process itself, which again is found to be heterogeneous across European countries. Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 166 Harald Sander and Stefanie Kleimeier II. Data In order to analyse central bank policy rates, money market rates, and commercial bank lending rates, monthly interest rates have been collected from the CD-ROM version of the IMF's International Financial Statistics (IFS) for all EU member countries from January 1985 to December 1998. As lending rates the rates listed in line 60p of the IFS have been used, the central bank discount rates listed in line 60 have been used as policy rates, and money market rates as listed in line 60b have been used. Exceptions to this sampling procedure were the following: Due to changes in central bank policy, rates from line 60a were used for France as of July 1989 and for the Netherlands as of January 1994. For Luxembourg, the national rates were used for lending rates but not for policy rates. Due to the monetary union between Belgium and Luxembourg, Belgian money market rates are the appropriate policy rates to be used for Luxembourg. If a series was not available on the IFS, the series has been obtained from Datastream. This applies to Austrian, Danish, and Swedish lending rates where Datastream's commercial bank prime lending rates have been used and to French money market rates, where one-month money market rates have been available. British central bank policy rates and Greek money market rates were missing on both, the IFS and Datastream and have thus not be included in our analysis. One should note that in particular lending rates are often heterogeneous across countries. Only recently, also the ECB has started to publish retail lending rates on a regularly base for EMU member countries, however, these data are also coming with the warning that these data are not fully harmonized. In interpreting any estimation results as evidence for heterogeneity across countries one should therefore bear in mind these limitations in the database. A detailed description of the data can be found in table A-l in the appendix. For analysing the impact of monetary policy on lending rates, there is an issue of what proxy for monetary policy to choose. Table 2 presents some basic correlation among interest rates in European countries. While in the whole period from 1985 to 1998 we find a varied picture of correlation between money market and policy rates, the sub-period3 from 1994 to 1998 shows in all cases, with the notable exception of Austria, a correlation close to one. This justifies in particular for the second subperiod to use money market rates as a proxy for the monetary policy 3 The later conducted structural break tests for the cointegration regression leads to the choice of this sub-period. Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 Asymmetric Adjustment of Commercial Bank Interest Rates in the Euro Area 167 stance as European central banks increasingly tried to exercise influence over the money market rate. This is in line with Bernanke and Blinder (1992) who favour the US-federal funds rate as a proxy for the monetary policy stance. When investigating whether the money market or the policy rate is more closely related to the lending rate the evidence is mixed. Basing the judgement on correlation coefficients, in the first subperiod one would favour the policy rate, while in the second sub-period the evidence clearly speaks in favour of the money market rate, with the exception of Finland and (to a much lesser degree) Germany and Italy. It therefore appears that a "discount rate addiction" (Cottarelli and Kourelis, 1994), i.e. the announcement effect of a discount rate change that induces banks to change their lending rates, has lost in importance as banks increasingly seem to base their pricing decision on cost of funds considerations. The latter observation is also reflected in the fact that basically in all cases the correlation coefficients between money market and lending rates have increased, with the notable exception of Germany4. In conclusion, we will concentrate here on the relationship between money market and lending rates5. III. Analysis of Symmetric and Asymmetric Adjustment of Lending Rates in Europe 1. The Pass-Through Approach Beginning with Cottarelli and Kourelis (1994), a growing literature is discussing the response of lending rates to monetary policy impulses as an important part of the monetary transmission process. These approaches typically model the transmission process in a dynamic model for the lending rate such as k* n* (1) U = Pi + Y, fa -i + + E fa a Mt-i+et i = 1 i = 1 where Lt and Mt are lending and money market rates, respectively, k* and 7i* are defined as the model's optimal lag-length. The estimated coef4 Mojon (2000) reports similar results (with respect to short-term credits to firms) when comparing the sub-periods 1979-1988 and 1988-1998. In his sub-samples all correlation increase except for Germany and Italy. In our sample division it becomes clear that Mojon's result with respect to Italy was basically due to the low correlation in the 1985-1993 segment. 5 The only exception is Greece where the money market rate is not available and is substituted by the policy rate. Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 168 Harald Sander and Stefanie Kleimeier - S Cd Oi « 2 2 ^ Ö a ^ in cd >4 O) ö o I> 00 O Tf Tf o i-H 05 00 Oi 05 OS O O O O O O 0505000505050500 OOOOOOOO CSI ^ CSj Oi o o CO CO lO ^ to tO O) ^ o o o ¡z; OiCOD-OOOCOC-OO COCONOICOOIMO) OOOOOOOO 00 CJ5 OS QJ rH ti 1 cd tjh PS w> 2 ö T3 Ö o PH COCDi-HOiCOCSllOI>OOOOOOOOiCO^ N0)00050)050)000)0)05000)05^ ooooöooooooooo^ Oi CO o CS] CT) CO D— CO lO 00 00 (M 1 CO lO 00 CO CT CT [> lO CT CO CT I> o o © o O O Ö o o O o O o ¡z; CO PC? V Oi 0 ^ dl 1 4S co § OS 05 >> 1—1 £ 1 Ö lO »S £ o o o o o o o o o o o < £ lO 00 o CO 00 t> CO o 00 m rH CT CO iq [> <N CT CT 00 CT 00 CT CS] o o o o o o 5 O o o o o O o o , o/ s - 3 i 5b g 'öl 0) PQ P ö <p id 5h CU CD in Ü Ü -H r—? >< <U "cd S3 Jh -1 —I I—I I—I Cd ö 00 a; 3 .S 'g S ft ^ PL, CO CO ¡3 Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 Asymmetric Adjustment of Commercial Bank Interest Rates in the Euro Area 175 Starting with the well known Engle and Granger (1987) symmetric cointegration model, the usual Dickey-Fuller (DF) and augmented DickeyFuller (ADF) tests have to be conducted. The DF tests are based on the estimated residuals of the cointegration regression (5) Aut = -60ut-i + et where the ¿-statistic for the estimated coefficient <$0 provides an indication regarding the cointegration of the two series. The ADF test is obtained from the regression where the optimal lag length c* is found based on the minimum Akaike information criterion (AIC) criteria for lags up to 12. According to Kremers et al. (1992) a residual-based test imposes a common factor restriction which lowers the power of the DF test procedure. In order to avoid rejecting the symmetric cointegration model too often, we therefore perform an additional t-test on (3ECT of equation (4) as in indicator for cointegration. However, if interest rates are not adjusting symmetrically to deviations from the long-run equilibrium, the Dickey-Fuller tests may fail to detect cointegration. Fortunately, a growing number of different specifications of asymmetric adjustment models are available. We consider five different specifications for asymmetric adjustment of interest rates. The first model we consider is the threshold autoregressive model (TAR0) developed by Tong (1983). The model makes a distinction whether the explained interest rate (lending rate in our case) is above or below its equilibrium level. Thus, the TAR0 allows for asymmetric adjustment depending on the sign of equilibrium-deviation. For example, if the money market rate decreases without an immediate adjustment in the lending rate, we obtain a positive realization of the error term ut. When in this case the autoregressive decay is faster than in the case of money market rate increases, then the lending rate adjustment is faster downward than upward. An appropriate test procedure is to set a Heaviside indicator It for different states of ut-\. c* (6) A ut = -So ut SiAut -i + et Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 176 Harald Sander and Stefanie Kleimeier Using this definition, we test for cointegration by estimating equation (8), which represents an modification of the ADF test. The null of no cointegration is rejected if the estimated F-statistic for H0 : pi = pi = 0 is significant based on critical values provided by Enders and Siklos (2000). 771* (8) AUt = ItPiUt-1 + (1 - It)p2Ut-l +Y1?2 + jAWt-i ¿ = 1 with the optimal lag length ra* determined via the minimum AIC criteria for models with up to 12 lags. When cointegration is established, an F-test for equality of pi and p2 indicates the presence of asymmetry. The second model (TAR*) is a modification of the TAR0 in the sense that the threshold that was formerly implicitly set at zero is now allowed to deviate from that value. The rationale behind such a non-zero threshold is that one or both variables may only adjust to a disequilibrium once it exceeds a certain minimum deviation in one direction. For example, the lending rate will adjust fast only when out of an equilibrium situation the money market rate drops in a way that the deviation from equilibrium exceeds an optimal threshold of, say, 0.5 percentage points. For lower deviations or increases in the money market rate, adjustment takes place at a significantly slower pace. Now the Heaviside indicator in conjunction with equation (8)11 is defined as r i if Ut-1>05 1 j i-l0 if ut-i < a*0 Following Chan's (1993), the optimal threshold a0* is found by searching over the mid-80 % of the distribution of ut and selecting the model for which the residual sum of squares is minimized. Cointegration and asymmetry testing proceeds with the above described F-tests. The third variation is a Band-TAR model (B-TAR*), which defines the Heaviside indicator as ( Ii = 1 if ut _ i > a*0 and 0 otherwise I2 = 1 if |wt_i|<aj and 0 otherwise h = 1 if ut-1 < -a*0 and 0 otherwise ii For both, the TAR* and the following B-TAR* model, the optimal lag length ra* of the TAR0 specification is used. Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 Asymmetric Adjustment of Commercial Bank Interest Rates in the Euro Area 177 while equation (8) has to be modified to 3 m (H) A Ut = ^ PjljUt - 1 + P3 +1 Awt _ i + £t j = 1 i = 1 Procedures for optimal lag length m* and optimal threshold a*0 are corresponding to those of the TAR* and the F-tests for cointegration and asymmetry are applied to all three coefficients pj. Such a model has often been applied in particular to model interest rate cointegration where infrequent and discrete adjustments in the rates occur (e.g. Balke and Fomby 1997, Baum and Karasulu 1998). For example, if deviations from equilibrium are small and will therefore not lead to an adjustment of the dependent interest rate, one may find no cointegration within a narrow band bordered by a*0 and -a*0 while outside this band cointegration and thus an error correction mechanism may be present. In the context of our study, such behaviour could be related to the "menu cost" argument of lending rate stickiness such that banks only adjust lending rates when deviations are sufficiently large. However, if it happens that inside the band cointegration is found but not outside, this could indicate that banks implicitly insure their customers against excessive deviations from equilibrium by smoothing the response of the lending rate. Finally, our fourth and fifth models are so-called momentum threshold autoregressive (M-TAR) models. Whereas in the TAR models the autoregressive decay always depends on the degree of deviation from equilibrium, one could also imagine situations where the adjustment speed depends on how fast the rates move away from or towards equilibrium. Enders and Granger (1998) therefore propose an M-TAR model where the Heaviside indicator depends as follows on the change in error correction term, Aut Similar to the TAR0 and TAR* specifications, the threshold in the M-TAR can either be set at zero leading to the M-TAR0 specification or be optimised at a*0 leading to the M-TAR* specification12. Cointegration and asymmetry testing proceeds based on equation (8) above. The M-TAR models have successfully been applied to the term structure of interest 12 When optimising the threshold in the M-TAR* model, the optimal lag length m* from the M-TAR0 is used. (12) Aut -1 > a0 < a0 Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 178 Harald Sander and Stefanie Kleimeier rates by Enders and Granger (1998) and Enders and Siklos (2000). According to the latter authors, M-TAR adjustment can be especially useful when decision makers (in our case banks) are viewed as attempting to smooth out large changes in a series. In sum, the analysis proceeds in seven steps: (1) Estimation of the cointegration relationship to obtain the estimated residuals ut. (2) Estimation of all five TAR-type models. (3) Cointegration test for each TAR-type model. (4) Asymmetry test for equality of coefficients in case of cointegration for each TAR-type model. (5) Repetition of steps (2) to (4) with varying lag-length to optimise AIC for each TAR-type. (6) Selection of optimal TAR-type model based on the minimum AIC across all model specifications. (7) Symmetric cointegration testing if step (6) can not establish asymmetric cointegration based on the optimal TAR-type model. The results of the cointegration analysis are summarized in Table 513. Looking first at the Engle-Granger model of symmetric cointegration, the Dickey-Fuller test statistics reveal that out of 15 countries there are five countries where no cointegration could be found and in three countries only marginal evidence for cointegration could be detected. When looking additionally at the t-tests for (5ECT these statistics indicate symmetric cointegration for two more countries and in two of the three marginal cases cointegration can be confirmed. However, in ten cases we find evidence for asymmetric adjustment. In such cases TAR-type models are better suited to detect cointegration. In particular, there is stronger evidence in favour of cointegration in the cases of Austria, Denmark, Finland, France, Italy and Spain where formerly there was conflicting, marginal, or no evidence for cointegration. In total, for these ten countries we selected an asymmetric cointegration model. Out of these five cases the M-TAR* model was most appropriate whereas the B-TAR* was selected three times and the TAR* twice. It is worth noticing that all asymmetric models selected are indicating that there exist non-zero !3 Results for the Engle-Granger cointegration regression are given in Table A-2. DW, DF and ADF tests are available from the authors upon request as are the results regarding the estimated coefficients and test statistics of equations (4), (8) and (11). Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 Asymmetric Adjustment of Commercial Bank Interest Rates in the Euro Area 179 43 tuo * 2 3 U £ 'S 1 » S CO w CO OH « « 2 ÉS « tf « « « ~ tó < < g g < 0 < Ë-H H ^ ^ M ^ H 2 2 CO CO 5 co 2 a s « a co E-i <J ti OT ri <*> o £ § & § 1/3 o £ § CO W M N N <u <D 2 2 Q o3 £ <u 0» di >> g ts g Ö « SP ° 3 p -O H fi v •rH W o Cd V X2 o o o o o O •c 'u "EH 'ÎH -M "S A> eu eu 0> s a e a B g a s e s o o B g >> >> ^ >> fi fi >» w w w w w CO cO cO co cO cO CO w w CO w" W C/Ì O> <D eu 0> O) (V >> >> >> >i s s >i cu >> PH O TH 00 OS CSI o CSI 00 C-co 00 Od <1 CO Cvi IH LO CSI RH CTI TJJ 00 OS OS ai 1—1 g LO oi 00 o 00 o O 00 co co I> co T-i iti 1 lO co CO (M co 1 Ci 1—1 in CM CSI t-H IO LO p-1 S 1 1 I—1 o PH I> (M ^ ^ 00 (M o co co 00 co 00 co i-H < CO CO CS] RH IO co 00 1—1 ^ 00 o O ^ CSI S i-H o i—I 00 ^ ^ LO csi 1-Î O csi ^ t> co O 1 CO CO LO t-H ai TH co co co CM IO IO (M S 1 1 t-H * (M (M o co co M 00 co co o co co ^ l-H I—1 CO oo TJJ co © co 1fi LO co in co Od lft co S CO LO o t> TH I> 00 TH O evi co evi 00 csi r-1 lO CO ^ CO co 1 OS co co (M t-H in ^ (M PQ 1 1 t-H CO o Ol 00 o co T-H co t-H IO Od o ai co PH iH LO 00 Ol if} CV] co t> 00 ^ co co t-H I> ai <« i-H csi 00 TP ce csi LO O csi o o in co O co H »ft 00 1 CO CO 1 co OS t-H co co co iH iH IO LO CSI CO (M t-H ^ CSI o o o 00 co CSI CSI IO co PH Ol LO Ol 00 OS OS RJN t-H 00 I> Q CSI LO O o LO t> co co csi LO O co 00 I> LO S CO CO 1 LO rH 1 O) T-l co co co t-H 1-H IO IO CSI ft fi CD R—L tuo 'co B a o CO > u T-H < fi s o u co •c C/3 e s g 'a; PQ fi co fi CO •S ÌH >> fi CO g S a> ÎH (1) O) Jh Ü Ü T3 fi co T3 fi co di £ CO tuo s t! o PH fi ai e fi 5 co co p M Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 180 Harald Sander and Stefanie Kleimeier thresholds, a finding that is in line with the "menu cost" argument. In the cases of Germany, the Netherlands, and Portugal no asymmetric cointegration was found but the adjustment process is basically symmetric in which case the Engle-Granger cointegration model applies. Finally, in the case of Greece and Sweden no cointegration and thus no error correction mechanism could be found. b) The Pass-Through of Money Market Rate Innovations in Europe Based on the cointegration testing we select the appropriate model for analysing the pass-through of interest rates as given in the last column of Table 5 and - for the sake of convenience - again in the second column of Table 6. In the case where no cointegration was found, we use the standard pass-through model (STD). This can be done by estimating the error correction model of equation (4) with (3Ect set to zero. For this as well as for all other specifications of equation (4), we have chosen an optimal lag length /c* and n* for lending and money market rates, respectively, by applying the minimum AIC criteria for all models with up to 12 lags in either interest rate. Consequently, in the STD model the impact multiplier is given by the estimated coefficient ¡32 and the longrun multiplier 6 is calculated according to equation (2). When cointegration was found, the long-run multiplier 0 is directly obtained from the cointegrating regression reported in Table A-2 of the appendix while again the impact multiplier is fa obtained from the appropriate specification of equation (4). The error correction mechanism itself depends on the optimal model selected in Table 5. In the case of the symmetric cointegration model (SYM), the ECT is equal to the estimated residuals of the cointegrating regression. (3Ect is therefore estimating the speed of a symmetric adjustment process towards a long-run equilibrium. In the models with asymmetric adjustment, 0Ect and the ECTs are 2-dimensional or, in the case of the B-TAR*, 3-dimensional vectors which give the speed of adjustment depending on the definition of the ECTs of equations (7), (9), (10), or (12), respectively. Furthermore, where appropriate, the value of the optimal threshold a*Q is given. Looking at the results of the pass-through analysis, we first can confirm the findings of earlier pass-through studies that within Europe the stickiness of the lending rate as measured by the impact multiplier varies considerably. It ranges from zero (i. e. an insignificant impact multiplier) in Austria, Denmark, Luxembourg, and the Netherlands up to Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 ASYMMETRIC ADJUSTMENT OF COMMERCIAL BANK INTEREST RATES IN THE EURO AREA 181 TUO CD J SH T3 - 2 CO O B a • H EFL CU u O TJH O 3 < M—I O CD A CO Ö O EH O RJ? EH U J? OS CO O O O T—I G O O ÎH W EH O D- © CO OS M TO CSI M CS] OS O 1-1 CO T-L ^ CO O IO O ^ T> CO T> O O O I I O O I I IO CSI CSI CO T-L ^ I I G OL) £ < § 3 J S CO O OS O OI 00 1—1 RH CO CSI OI 1-H O O CO 1-T 1—1 CSI OS © IO © OS © © Ö © 1-1 Ö 1—1 © © 1—1 £ A ^ A £ ^ â 3 S CO TJH I> CSI I> O CTS OL ^ CO I-H CO © 1-H H [> © T-I © T-I CSI CS] O ^ LO CO 00 O CO OS CO CCS] CD "TI DCO O CSI O CO O CO <v O S PS S * « * « < < EH EH S 2 « S I PQ Q EH Cn PC! £ I PQ Q H CO Ö O U CO S W < S BP "A; PQ CO S Ö CU A T3 TÌ JS 15 £ >> Ö CO S f-i <v o o TJ Ö CO KREDIT UND KAPITAL 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 182 Harald Sander and Stefanie Kleimeier * SS w tUO cd iJ * Optimal Threshold a0* -0.065 -0.368 0.141 -0.203 : Adjustment PECT, 3 0.064 (0.241) 0 T5 CU «M DH U « «s 0 '-M -0.103 (-2.270) -0.660 (-1.312) 0.715 (1.393) -0.439 (-2.538) Error Corree PECT, 1 0.053 (1.295) -0.170 (-2.281) -0.222 (-2.447) -0.156 (-2.494) -0.071 (-0.438) -0.103 (-0.893) -0.176 (-2.709) Long-run Multiplier e 0.866 0.475 0.083 0.983 1.170 1.074 1.074 0.698 0.636 Impact Multiplier & 0.197 (7.027) 0.084 (1.058) 0.089 (1.114) 0.115 (0.934) 0.168 (2.214) 0.705 (6.169) 0.735 (6.543) 0.552 (2.589) 0.272 (4.923) Model M-TAR* M-TAR* STD SYM SYM B-TAR* SYM STD M-TAR* Country Italy Luxembourg Netherlands Portugal Spain Sweden United Kingdom Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 Asymmetric Adjustment of Commercial Bank Interest Rates in the Euro Area 183 levels above 0.7 in Belgium and Spain. It is also interesting to note that our estimates fall by and large in the range indicated in the earlier studies reported in Table 3 with the only exception of the Netherlands. However, in this country we find a reasonably strong symmetric error correction mechanism at work that brings the lending rate back to its equilibrium relationship with the money market rate. The long-run multipliers are close to one (full pass-through) in most but not all cases. In Austria, Denmark, France, Greece, Luxembourg, Sweden, and the UK there is no full pass-through even in the long run. Looking at the different optimal models it is striking how diverse the workings of the banking sector in Europe still are at the advent of the single currency. For example, in Greece and Sweden no error correction mechanism could be identified. In these countries the impact multipliers appear to be quite high but there is no full pass-through in the long run. On the other hand, we find symmetric adjustment processes in the cases of Germany, the Netherlands, Portugal, and Spain, although in the latter case the speed of adjustment was found insignificant. A threshold autoregressive adjustment was found in five countries. Austria and France are characterised by a TAR* adjustment process. In Austria there is a strong indication that an error correction process is only invoked if the lending rates are above their equilibrium level by more than the 1.577 percentage points indicated by the optimal threshold. This "shielding" of the customers may have been responsible for not finding cointegration in the Engle-Granger model. By contrast, in France the threshold is lower and the test statistics indicate a close-to-nonexistent adjustment process. The B-TAR* models that have been selected for Finland and Ireland show significant error correction mechanisms. Since the B-TAR* was marginally significant for Spain, we additionally report this model. In Finland it surprisingly appears that even small deviations within the band will lead to adjustments (see PECT,2 = -0.134) as well as situations where the lending rate is below its equilibrium level (see PECT,3 = -0.168). Ireland exhibits for all three error correction mechanisms a full or even overshooting adjustment with coefficients close or even below -1. However one should recall that we had problems establishing the 1(1) property for Irish interest rates. Therefore, we also report the STD model. M-TAR* models have been advocated for cases in which increasing or decreasing deviations from equilibrium may induce strong and asymKredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 184 Harald Sander and Stefanie Kleimeier metric reaction. For Belgium, positive discrepancies from long-run equilibrium resulting from a decrease of the money market rate or an increase of the lending rate such that the Aut < -0.233 are eliminated very quickly (see PECT,I = -0.257) whereas other changes display a large amount of persistence (see (3ECT,2 = -0.043). A similar pattern can be found for Luxembourg, which could be expected under the monetary union between the two countries. The opposite appears to be the case for Denmark and Italy. In the UK the adjustment mechanism on both sides of the threshold is significant but the speed of adjustment is faster for negative discrepancies. IV. Summary and Conclusion Our study extends the traditional pass-through literature by incorporating an error correction mechanism that is based on cointegration analyses allowing for symmetric as well as for a variety of asymmetric adjustment mechanisms. By and large the results of earlier pass-through studies are confirmed, in particular the finding that monetary policy in euro area is still to be conducted under the conditions of an "asymmetric EMU" of which the differences in the way the different banking systems in euro area countries work are arguably among the most important ones. However, not only is the speed of adjustment different across countries but as we find that on a more fundamental level the nature of the adjustment process itself is heterogeneous. Therefore, our analysis provides a deeper insight into the differential workings of the banking markets across Europe than previous studies. While optimists hope that the elimination of currency risks may contribute to an institutional harmonization within EMU the evidence provided here suggests that for the nearer future asymmetries will continue to influence the monetary mechanism within the euro area. However, with an increasing knowledge of the degree of heterogeneity in European banking markets, shooting at a moving target might become an easier task in the future. Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 Asymmetrie Adjustment of Commercial Bank Interest Rates in the Euro Area 191 tober. - Kashyap, A. N. and Stein, J. C. (1993): Monetary policy and bank lending, NBER Working Paper 4317, April. - Kieler, M. and Saarenheimo, T. (1998): Differences in monetary policy transmission? A case not closed, European Commission Directorate-General for Economic and Financial Affairs, Economic Papers 132. - Kleimeier, S. and Sander, H. (2002): Consumer credit rates in the Eurozone, European Credit Research Institute (ECRI), Brussels, ECRI Research Report No. 2. - Kleimeier, S. and Sander, H. (2000): Régionalisation versus globalisation in European financial market integration: Evidence from co-integration analyses, Journal of Banking and Finance 24(6), 1005-1043. - Kremers, J. J. M., Ericsson, N. R. and Dolado, J. J. (1992): The power of cointegration tests, Oxford Bulletin of Economics and Statistics 54(3), 325-348. - Modigliani, F. (1963): The monetary transmission mechanism and its interaction with real phenomena, Review of Economics and Statistics 45(1), 79-107. - Mojon, B. (2000): Financial structure and the interest channel of the ECB monetary policy, ECB working paper 40, November. - Quandt, R. (1960): Tests of the hypothesis that a linear regression system obeys two separate regimes, Journal of the American Statistical Association 55, 324-330. - Ramaswamy, R. and Sloek, T. (1997): The real effects of monetary policy in the European Union: What are the differences?, IMF Working Paper 97/160. - Sander, H. and Kleimeier, S. (2000): Asymmetric adjustment of commercial bank interest rates in the euro area: Implications for monetary policy, working paper presented at the 3rd Conference of the Swiss Society For Financial Market Research, Zurich, April 7, 2000 and at the "Financial Structure, Bank Behaviour and Monetary Policy in the EMU" conference, Groningen University, October 5-6, 2000, (mimeo). - Scholnick, B. (1996): Asymmetric adjustment of commercial bank interest rates: evidence from Malaysia and Singapore, Journal of International Money and Finance 15, 485-496. - Scholnick, B. (1999): Interest rate asymmetries in long term loan and deposit markets, Journal of Financial Services Research 16(1), 5-26. - Stiglitz, J. and Weiss, A. (1981): Credit rationing in markets with imperfect information, American Economic Review 69, 339-345. - Tong, H. (1983): Threshold models in non-linear time series analysis, Springer Verlag, New York. - Toolsema, L. A. and de Haan, J. (2000): Convergence of monetary transmission in EMU. New Evidence, Department of Economics, University of Groningen (mimeo). - Toolsema, L. A., Sturm, J.-E. and de Haan, J. (2001): Convergence of monetary transmission in EMU. New Evidence, CESifo Working Paper No. 465, Munich, April. - Wickens, M. R. and Breusch, T. S. (1988): Dynamic specification, the long-run and estimation of transformed regression models, The Economic Journal 98, 189-205. Summary Asymmetric Adjustment of Commercial Bank Interest Rates in the Euro Area: An Empirical Investigation into Interest Rate Pass-Through Our study extends the traditional pass-through literature by incorporating an error correction mechanism that is based on cointegration analyses allowing for structural breaks and symmetric as well as for a variety of asymmetric adjustment mechanisms. While some results of earlier pass-through studies regarding a symmetric monetary transmission mechanism within the euro area are confirmed, our Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38 192 Harald Sander and Stefanie Kleimeier study provides additional evidence that not only the speed of adjustment differs but that the nature of the adjustment process itself is heterogeneous across countries. Therefore, our analysis provides a deeper insight into the differential workings of the banking markets across Europe. (JEL E43, E52, E58, F36) Zusammenfassung Asymmetrische Anpassung von Geschäftsbankenzinsen im Eurogebiet: Eine empirische Analyse des Zins-Pass-Through Unsere Studie erweitert die traditionelle Literatur des Zins-Pass-Through durch die Berücksichtigung von Kointegrationsbeziehungen und der korrespondierenden Fehlerkorrekturmechanismen. Dabei erlaubt unser Ansatz, sowohl Strukturbrüche als auch symmetrische und verschiedene asymmetrische Anpassungsmechanismen zu analysieren. Während einige Ergebnisse früherer Pass-Through-Studien bezüglich des Transmissionsmechanismus innerhalb des Eurogebiets bestätigt werden, zeigt unsere Studie darüber hinaus, daß sich im Ländervergleich nicht nur die Anpassungsgeschwindigkeiten unterscheiden, sondern auch, daß der Charakter der Anpassungsprozesse selbst heterogen ist. Unsere Ergebnisse erlauben daher einen tieferen Einblick in die unterschiedlichen Funktionsweisen der Bankenmärkte Europas. Résumé Ajustement asymétrique des taux d'intérêt des banques commerciales dans la zone euro: une investigation empirique sue la répercussion des taux d'intérêt Notre étude porte plus loin la littérature traditionnelle en incorporant un mécanisme de correction d'erreurs basé sur des analyses de cointégration, tenant compte de ruptures structurelles ainsi que de mécanismes d'ajustements symétriques et asymétriques. Alors que certains résultats d'études passées sur un mécanisme de transmission monétaire symétrique au sein de la zone euro sont confirmés, notre étude montre une évidence supplémentaire: non seulement la vitesse d'ajustement diffère, mais aussi la nature du processus d'ajustement lui-même est hétérogène selon les pays. Notre analyse offre donc un aperçu plus profond sur les fonctionnements différentiels des manchés bancaires européens. Kredit und Kapital 2/2002 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.35.2.161 | Generated on 2023-01-16 13:18:38