Bank interest rate pass-through:
an empirical study on Germany and Spain
Rosa Keijsper (10624783) University of Amsterdam, faculty of Economics and Business Specialization: Economics and Finance Thesis supervisor: Mr. S. Singh January 2017 Abstract: The efficiency of the monetary policy of the European Central Bank depends highly on how much of the changes in the market interest rate are passed-through to the retail banking interest rates. This pass-through is affected by, inter alia, the level of competition in the banking sector in a country. This paper examines the interest rate pass-through for Germany and Spain, and compares them in terms of speed and completeness. Data from 2008 until 2016 is used, and the speed and completeness are estimated by the regression of an Error Correction Model. The results show a higher speed of adjustment for Germany than for Spain, but are less uniform for the completeness estimates. More research is needed for these specific countries, and for the reliability of the Error Correction Model as such.Keywords: Interest rate pass-through, monetary transmission mechanism, European Central Bank, bank interest rates
Statement of Originality This document is written by Rosa Keijsper, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of contents 1. Introduction 1 2. Literature review 2 2.1. Interest rate pass-through 2 2.2. Previous studies 3 3. Research question 5 4. Data and methodology 7 4.1. Data 7 4.2. Methodology 9 4.2.1. Single Equation Error Correction Model 9 4.2.2. Augmented Dickey-Fuller test for Unit Root 10 4.2.3. Engle-Granger Augmented Dickey-Fuller test for cointegration 11 5. Results 12 5.1. Augmented Dickey-Fuller test for Unit Root 12 5.2. Engle-Granger Augmented Dickey-Fuller test for cointegration 13 5.3. Single Equation Error Correction Model 15 5.3.1. New business loans to non-financial corporations 16 5.3.2. New loans to households for consumption 17 5.3.3. New loans to households for house purchases 19 5.3.4. New deposits from private households 20 6. Conclusion 22 References 24 Appendix 27
1. Introduction
Between October 2008 and May 2009 the European Central Bank (ECB) lowered the interest rate on its main refinancing operations by 325 basis points (Hristov et al., 2014). They did this in response to the start of the financial crisis, to stimulate nominal spending and investing. By now, January 2017, the ECB has lowered the interest rate on their main refinancing operations to 0%. For the rate on the deposit facility they went even further to a negative number of -0.4% (ECB, 2016).
The ECB cannot directly influence the economic output, but it has to rely on its monetary policy channels. One of their most important ways to affect the economic output is via the interest rate channel. When the ECB lowers their interest rate, they make it possible for banks and other financial institutions to borrow at a lower rate. They try to encourage banks to increase their lending activities, and lend at this lower rate as well. This is one of their ways to stimulate the output of the economy. The strength of this interest rate channel therefore depends on how much of these changes in the main refinancing rate are passed-through to the retail rates set by the banks, which is called the interest pass-through.
Literature shows that the banks in the Euro area used to have a relatively fast and complete pass-through (ECB, 2009). However more recent data shows that these banks now only partly pass-through these lower interest rates to their rates for loans and deposits (Hristov et al., 2014). Hristov et al. also find heterogeneity between countries in the Euro area. Therefore it is interesting to research the interest rate pass-through for individual countries, and try to find where these differences come from.
The degree of competition in the banking sector in a country is one of the main factors that influence the pass-through of interest rates (Mishkin et al., 2013). Therefore, this paper will focus on two countries where this degree of competition differs greatly, namely Germany and Spain. The main research question will therefore be: what is the difference between the interest rate pass-through effect in Germany versus Spain in the timeframe of September 2008 until October 2016, looking at the speed and the completeness of the pass-through, and calculated using a Single Equation Error Correction Model (ECM)?
The remainder of this paper will try to answer this question and is organised as follows. Section 2 will give the necessary information to understand the topic of this
will give a further motivation of the research question. Section 4 will include a description of the data as well as the exact model that will be used. Next, in section 5 the empirical results will be presented, discussed and compared to previous literature. Lastly, the paper will be concluded in section 6, together with some final remarks and suggestions for further research. 2. Literature review This section will begin with an explanation of the interest rate pass-through and factors that affect the degree of speed and completeness. Following, there will be a discussion of previous studies and their results. 2.1 Interest rate pass-through Keeping prices stable with an inflation rate of approximately 2 percent is the main goal of the European Central Bank. The ECB can influence prices via their decisions on monetary policy, which are then slowly transmitted through the real economy. The channels via which a central bank can influence these prices are called Monetary Policy Transmission Mechanisms (MPTM). The ECB shows that there are four different channels; the money/credit channel, the asset price channel, the interest rate channel and the exchange rate channel (Mishkin et al., 2013). One of the most important MPTM’s via which the ECB can influence the inflation is the interest rate pass-through. This is especially the case in the Euro area, because here the financial system is mainly bank-based (Andries & Billon, 2015).
The interest rate pass-through is defined as: “the adjustment of retail bank interest rates in response to changes in the official policy rates set by the central bank” (Mishkin et al., 2013, p. 607). The main idea of the interest rate pass-through is that central banks influence the money markets and thus the money market rates. Changes in the money market rates then influence lending and deposit rates from retail banks with different degrees1. The decisions that banks make regarding the interest rates that are paid on these loans and deposits then influence the expenditure and investment behavior of the deposit-holders and borrowers. In this way the central bank can
1 Reasons for this are; differences in elasticity of demand and supply, important switching costs in the market for deposits and risky borrowers in certain market segments (Andries & Billon, 2015). A further explanation of the differences between certain rates can be found in section 2.2.
influence real economic activity. Depending on how fast and more complete this pass-through is, determines the influence that the central bank has (De Bondt, 2005). However, the pass-through is not always complete. There are several reasons, which will be explained further.
Incomplete pass-through
One of the main reasons for an incomplete pass-through is imperfect competition amongst banks (Mishkin et al., 2013). The explanation is that banks in a competitive banking sector will change their lending and deposit rates more quickly if the policy rate decreases, since they prefer to avoid a loss in market share. However, if there were less competition amongst banks, keeping your loan rates up would increase your profit margin, without the chance of a decrease in market share.
Nonetheless, even in a market with perfect competition there is not always a complete interest rate pass-through. This is because of market imperfections, which lead to banks having a certain level of market power. Two of these market imperfections are the existence of switching costs (De Bondt, 2005) and the degree of monetary policy transparency (Kleimeier & Sander, 2006). Moreover, Hristov et al. (2014) mention some frictions that banks face that lead to stickiness of interest rates, and thus incompleteness or slowness of this MPTM. Those frictions are: a high cost of adjusting interest rates, tight collateral restrictions and high costs of restoring a banks’ capital position after a shock. Lastly, Sander and Kleimeier (2004) show that certain structural features of national financial markets have a considerable effect on the pass-through. Moreover, they show that the degree of growth, and legal and cultural differences between countries also affect the completeness of the pass-through. This therefore explains the differences in the pass-through between countries in the Euro area. 2.2 Previous studies The interest rate pass-through has been widely researched since the 1980s. Most of the studies pay attention to the heterogeneity of the pass-through effect looking at the speed and the degree of completeness. However, the literature differs across the econometric models used, interest rates selected and countries that are chosen to conduct the research on. For example, there are papers written about individual countries such as
(De Graeve et al., 2007), Portugal (Rocha, 2011) and Australia (Lowe & Rohling, 1992). However, there has also been extensive research on the whole Euro area (Andries & Billon, 2015; Sorensen & Werner, 2006; Hristov et al., 2014; ECB, 2009). Furthermore, the interest rates that are used differ amongst studies. Different loans and deposits rates are used with different maturities for the retail bank rates, and several different rates are used as market rates (for example: EURIBOR, EONIA, Money Market Rate).
In addition, there have also been many papers discussing the effect the financial crisis had on the completeness and speed of the pass-through (Hristov et al., 2014; Karagiannis et al., 2010; ECB, 2009). Before the financial crisis there were differences in the completeness and speed of this effect between banks and countries within the Euro area (Kleimeier & Sander, 2006). Nonetheless, in the time before the financial crisis this pass-through was seen as ‘generally complete’ in the Euro area (Hristov et al., 2014). Moreover, in 2006 researchers thought that it would become faster and more homogenous over the years (Kleimeier & Sander, 2006). However Hristov et al. also showed that, since the financial crisis in 2008 the pass-through became distorted and this lowered the effect of monetary policy conducted by the ECB.
Methodology
There are two main econometric models used in previous research. First, the single-equation error-correction model (ECM) was introduced (Engle & Granger, 1987). An ECM estimates the speed at which the dependent variable returns to equilibrium after a change in one of the independent variables. This model helps to calculate both the short and the long-run pass-through and the degree of completeness. Later, a vector error-correction model (VECM) was introduced by Johansen (1991), which adds error correction features to the Vector Autoregressive Model (VAR). The VECM allows for interdependence between variables.
Results
Generally the literature about interest rate pass-through shows agreement on the speed and completeness of the effect. For example, studies show a higher long-run pass-through on lending rates than on deposit rates (De Bondt, 2005; Kleimeier & Sander, 2006). This is explained by the argument that the elasticity of demand for loans is bigger than the elasticity of demand for deposits. Moreover, this long-run pass-through
on deposits is found to be more complete for longer maturities than for short maturities (De Graeve et al., 2007; De Bondt, 2005). De Bondt (2005) argues that this may come because of high switching costs in this market. Furthermore, the long-run pass-through is found to be more complete than the short-run pass-through. Cottarelli and Kourelis (1994) say that this may come because in financial markets banks can switch to a different source of financing than loans. However, in the short-run switching between sources does not outweigh the costs of it, whilst in the long-run it does. Moreover, in the short run there is more information asymmetry, which disappear over time. This is another reason why the long-run pass-through might be more complete than the short-run one. Also in the short-run pass-through there are differences among bank products. For example, the short-run pass-through is found to be higher for interest rates on deposits with a longer maturity, than for a shorter maturity (De Graeve et al., 2007). And it is higher for lending rates to enterprises than to households (De Grave et al., 2007; De Bondt, 2005).
Nonetheless, there are also many contradictions found in the literature. For example some researches find an incomplete pass-through in the long-run (De Graeve et al., 2007; Sander & Kleimeier, 2004), whilst others find a more complete pass-through in the long-run (De Bondt, 2002). These contradictions arise because of the usage of different econometric models, countries, time periods and interest rates.
Concluding, there has been a lot of research on the interest-rate pass-through, and also exclusively on the pass-through since the 2008 financial crisis in the Euro area. Overall, the crisis has led to a decrease in the effect of this MPTM, and the heterogeneity in the Euro area shows that the crisis affected this MPTM differently in different countries. As explained before, the level of competition in the banking sectors is one of the main influencers of the interest rate pass-through.
3. Research question
Altogether, this has lead to the question if countries with less frictions in the competitiveness in the banking sector (like the Germany) show a smaller level of speed and completeness of their interest rate pass-through, than countries with more of these frictions (for example Spain). Therefore my research will further investigate if this is the case.
The structural financial indicators published by the Statistical Data Warehouse in 2016, show evidence that indeed the Spanish banking sector is less competitive than that of Germany. For example, since 2011 the share of total assets of the five largest credit institutions in Spain (a peripheral country) grew from 48.1% to 60.2%, whereas for Germany (a core country) it remained stable around 32% (see Table 1). Moreover, the ECB also shows that since the financial crisis Spain reported the third largest decrease in the number of credit institutions in the whole Euro area (ECB banking structures report, October 2014). Table 1: Share of total assets of five largest credit institutions (in percent) 2011 2012 2013 2014 2015 Germany 33.5 33.0 30.6 32.1 30.6 Spain 48.1 51.4 54.4 58.3 60.2 Note: adapted from Structural financial indicators, Statistical Data Warehouse, 2016. Lastly, there is a different way to calculate the concentration of markets, namely via the Herfindahl-Hirschman Index (HHI). The HHI is defined as “the sum of the squares of the market shares of all firms within the industry, where the market shares are expressed as fractions” (ECB banking structures report, October 2014, p. 14). In this way one can calculate the concentration and thus the competitiveness of the banking sector. This index ranges from 0 to 10,000, with 0 being a very competitive sector and 10,000 being a sector with only one player. As can be seen in Table 2, the German banking sector went from 317 in 2011 to 273 in 2015, which shows that Germany became more competitive. The Spanish banking sector, however, went from 596 in 2011 to 896 in 2015. This shows that the Spanish banking sector was less competitive to begin with, and that the gap between Germany and Spain is growing even further. Table 2: Herfindahl index for credit institutions (index ranging from 0 to 10,000) 2011 2012 2013 2014 2015 Germany 317 307 266 300 273 Spain 596 654 719 839 896 Note: adapted from Structural financial indicators, Statistical Data Warehouse, 2016.
4. Data and methodology 4.1 Data
This section will begin with an explanation of the retail and market interest rates that are used in this research. Following, the time frame will be explained and lastly there will be some more general information (number of observations, etc.) on the data.
I use two kinds of interest rates: the market rate that is set by the ECB, and retail rates set by banks in Germany and Spain. Several rates from retail banks in Germany and Spain are used, which will be split up into loans and deposits. These rates will also be split up into maturities of shorter than one year and longer than one year. In the remainder of this paper they will be referred to as short-term and long-term rates respectively. An overview of the retail rates that will be used are shown in Table 3 on the next page. These rates can all be found for both Germany and Spain on the Statistical Data Warehouse of the ECB. The rate that the ECB has direct control over is called the interest rate on main refinancing operations (MRO). However, since the MRO changes infrequently, using this rate will cause for imprecise econometric estimates (De Bondt, 2005). Therefore studies on the pass-through subject use different interest rates as a proxy for this MRO. Aristei and Gallo (2014) recommend using the 3-month EURIBOR (Euro Interbank Offered Rate). The EURIBOR rate is based on the average interest rate at which a panel of the largest banks in the Eurozone make unsecured loans to each other (Mishkin et al., 2013). So the EURIBOR is the rate the banks charge to each other for borrowing and lending, whilst the main refinancing rate is the rate at which banks in the Euro area can borrow at the ECB. Abbassi and Linzert (2012) argue that because the EURIBOR incorporates expectations on future short-term interest rates it indeed influences how well the transmission mechanism in the Euro area works. Because of these arguments, in this paper the 3-month EURIBOR will be used as a proxy for the market rate.
The data that will be used starts in September 2008. This is because this is the month that the Lehman Brothers filed for bankruptcy. Since this started a lot of turmoil in the financial markets, this month is often referred to as the beginning of the international banking crisis. After this date the ECB drastically lowered their interest rates. Therefore the data for this research will start in September 2008. Moreover, the data that is needed is available until October 2016.
Table 3: Overview of retail bank loans and deposits used Loans Deposits Interest rates on new business loans to non-financial corporations (all) (NFC) Interest rates agreed on new deposits from private households (all) (DEP) Interest rates on new business loans to non-financial corporations (up to 1 year) (NFC <1) Interest rates agreed on new deposits from private households (up to 1 year) (DEP <1) Interest rates on new business loans to non-financial corporations (over 1 year) (NFC >1) Interest rates agreed on new deposits from private households (over 1 year) (DEP >1) New loans to households for consumption (all) (CON) New loans to households for consumption (up to 1 year) (CON <1) New loans to households for consumption (over 1 year) (CON >1) New loans to households for house purchases (all) (MOR) New loans to households for house purchases (up to 1 year) (MOR <1) New loans to households for house purchases (over 1 year) (MOR >1) All rates are annualized agreed rates, which means that these are the rates agreed upon between the bank and the client, converted to an annual basis and quoted in percentages per annum. They cover all the interest payments, but no other charges (ECB, October 2003). Moreover, monthly data is used, because this is the shortest interval for the retail interest rates given by the ECB and it gets rid of the daily spikes of the EURIBOR. Considering that there are no rates missing in the data, this gives 98
observations per interest rate.
4.2 Methodology This section will begin with an explanation of the main model used in this paper: a Single Equation Error Correction Model. Following, a test for stationarity of interest rates will be explained and lastly a test for cointegration is discussed. 4.2.1 Single Equation Error Correction Model A single equation error correction model (ECM) will be used, because with this one can determine the speed and completeness of retail interest rates to changes in market interest rates set by the central bank (Sorensen, & Werner, 2006).
The specific single equation ECM proposed by Sorensen and Werner (2006) is displayed in Equation (1).
∆𝐵𝑅!,! = 𝛾 + 𝜃 𝐵𝑅!,!!!− 𝛽 𝐸𝑈𝑅𝐼𝐵𝑂𝑅!!! + 𝛼!∆𝐸𝑈𝑅𝐼𝐵𝑂𝑅!+ 𝛼!∆𝐸𝑈𝑅𝐼𝐵𝑂𝑅!!!+
𝛿∆𝐵𝑅!,!!!+ 𝜀! (1)
Where changes in German and Spanish retail bank rates are explained by changes towards the long run equilibrium between this retail bank rate and the EURIBOR. This paper will focus on three most important estimates; the speed of adjustment (θ), the immediate pass-through (𝛼!) and the final pass-through (β).
The slope coefficient, θ, shows the speed of adjustment towards this equilibrium. When the retail rate moves toward the equilibrium on average, θ will be negative. This means that if there would be a positive deviation from the equilibrium in the past period this would lead to a negative change in the bank retail rate in this period. The size of θ shows the speed of adjustment towards the equilibrium. The bigger (negative) the number, the faster the bank interest rate moves back to the equilibrium after a change in the EURIBOR.
Moreover, parameter 𝛼! measures the immediate pass-through. It shows how much of a change in the EURIBOR is passed through to the retail rate in the short-run. When the number is close to one it means a complete pass through, and when it is significantly lower than one it shows that in the short-run there is a certain ‘stickiness’ in the pass-through.
The variable 𝛽 measures the final pass-through of the market interest rate to the retail interest rate. A 𝛽 of one means that in the long run a 1% change in the EURIBOR leads to a 1% change in the long run equilibrium of the retail rate. This would mean that
the long run pass-through would be complete. If the final pass-through is significantly lower than one, this indicates an incomplete interest rate pass-through.
Changes of past bank rates are added to avoid any model misspecifications (Sorensen, & Werner, 2006). The error term, 𝜀!, includes all the factors that cause the difference between the ith average score and the value that is predicted by the regression line. This error term contains all the possible omitted variables that also determine the value of ∆𝐵𝑅!,!. To regress this ECM it has to be rewritten to the form of Equation (2). ∆𝐵𝑅!,! = 𝛾 + 𝜃𝐵𝑅!,!!!+ 𝜌𝐸𝑈𝑅𝐼𝐵𝑂𝑅!!!+ 𝛼!∆𝐸𝑈𝑅𝐼𝐵𝑂𝑅!+ 𝛼!∆𝐸𝑈𝑅𝐼𝐵𝑂𝑅!!!+ 𝛿∆𝐵𝑅!,!!!+ 𝜀! (2) Where 𝜌 stands for −𝜃 ∗ 𝛽. After doing the regression, this model can be rewritten into the initial single equation ECM again. After doing the regressions one can compare the estimates of Germany and Spain to answer the research question if there are differences between the speed and completeness of the pass-through between the two countries. Where it is expected that a country with more competition in the banking sector (Germany) will show a higher speed of adjustment and a higher completeness than one with less competition (Spain).
Moreover, this ECM relies on some assumptions. Namely, that the interest rates that are used do not return to their old values (they are non-stationary) and the retail rates and the EURIBOR have a certain level of cointegration. These two things will be explained in the next sections.
4.2.2 Augmented Dickey-Fuller test for Unit Root
Literature indicates that the Single Equation ECM relies on the assumption of non-stationarity of interest rates. This means that the interest rates do not return to their old value (ECB, 2009). When a variable is non-stationary it has a unit autoregressive root, also called a unit root. Nonetheless, some papers do not even mention this subject, whilst others just assume interest rates to be non-stationary (Andries & Billon, 2015). However, in this paper the interest rates will be tested for non-stationarity to be sure the ECM assumption is satisfied.
The most common way to test for a unit root is to use the Augmented Dickey-Fuller (ADF) test (Stock & Watson 2015). The null hypothesis of the ADF test is that the interest rate has a unit root/is non-stationary. Under the alternative hypothesis the
interest rate does not have a unit root, so it is stationary. The OLS t-statistic is called the ADF statistic. The ADF test is an extension to the original Dickey-Fuller test, because it adds the lagged terms of the variable of interest. Using the ADF test circumvents the possibility of serial correlation (Stock & Watson 2015). The number of lags, p, that need to be used can be calculated with a formula given by Schwert (1989). Equation (4) displays this formula. 𝑝!"# = [12 ∗ (!""! )!!] (4) Because this study uses 98 observations, there will be 11 lags used in the ADF test. 4.2.3 Engle-Granger Augmented Dickey-Fuller (EG-ADF) test for cointegration When the assumption for non-stationarity is satisfied, we can estimate the short-term pass-through. However, to be able to estimate the long-term pass-through there is also a second assumption that needs to be satisfied; the interest rates should have a cointegration relation with the market rate. This means that there is a long-run equilibrium between the retail rate and the EURIBOR to which both rates revert. Stock and Watson (2015) propose the use of the Engle-Granger Augmented Dickey-Fuller (EG-ADF) test for cointegration. This test consists of two steps. In the first step the cointegrating coefficient, θ, is estimated by an OLS estimation of regression (5). 𝐵𝑅!,! = 𝛼 + 𝜃𝑋𝐸𝑈𝑅𝐼𝐵𝑂𝑅!+ 𝑧! (5)
If the two rates are indeed cointegrated, this regression will show the long-run equilibrium between the rates. α will show how much the bank rate is higher or lower than the EURIBOR rate on average. It is a constant and thus a stationary variable. However, 𝐵𝑅!,!− 𝜃𝑋𝐸𝑈𝑅𝐼𝐵𝑂𝑅! will result in a stationary variable only if the two rates
are cointegrated. If this is the case, there will be a linear combination of stationary variables, which will result in a stationary variable as well. To test this, an Augmented Dickey-Fuller t-test is performed in the second step to test for a unit root in the residual from the regression; 𝑧!. The null-hypothesis in this test is that the residual is non-stationary, which implies no cointegration between the two rates. Whereas the alternative hypothesis states that the residual is stationary and does imply cointegration. Yet, the critical values to be used in this step are no longer the same
as provided by Dickey-Fuller. Hamilton (1994) came up with the right critical values. In this case the critical value for a 5% significance level is -2.76.
5. Results
In this section the results for the stationarity test, the cointegration test and the ECM will be given, interpreted and compared to previous literature. The results for the ECM will be divided into the four different categories of loans and deposits.
5.1 Augmented Dickey-Fuller test for Unit Root
The Augmented Dickey-Fuller test for unit root was performed for each bank interest rate as well as the EURIBOR. The results are summarized in Table 4, on the following page.
This shows that for almost all interest rates the null-hypothesis of having a unit root (and thus being non-stationary) is not rejected. Only for the interest rate on new loans to households for consumption in Spain it can be rejected, but only at the 10% significance level. This is in line with the literature, that interest rates are indeed non-stationary, and thus do not move back automatically to their old value.
It is known that when a time series is non-stationary, it can be differenced to render it stationary (Stock & Watson, 2015). This is called “First-difference stationary”. For example, if 𝑌! = 𝛽!+ 𝑌!!!+ 𝑢! is non-stationary, then taking the first difference ∆𝑌!= 𝛽!+ 𝑢! is stationary. This is why the specific ECM that is used for the pass-through of interest rates contains the first differences of the interest rates.
Table 4: Augmented Dickey-Fuller unit root test
Interest rate H0 ADF test statistic P-value
NFC – GE Unit root -1.692 0.435 NFC<1 – GE Unit root -1.652 0.456 NFC>1 – GE Unit root -0.853 0.803 CON – GE Unit root -1.086 0.721 CON<1 – GE Unit root -1.625 0.470 CON>1 – GE Unit root -0.546 0.883 MOR – GE Unit root -0.711 0.844 MOR<1 – GE Unit root -1.086 0.721 MOR>1 – GE Unit root -0.703 0.846 DEP – GE Unit root -0.881 0.794 DEP<1 – GE Unit root -0.631 0.864 DEP>1 - GE Unit root -0.480 0.896 NFC – SP Unit root -1.073 0.726 NFC<1 – SP Unit root -1.329 0.616 NFC>1 – SP Unit root -0.667 0.855 CON – SP Unit root -2.820*** 0.056 CON<1 – SP Unit root -2.332 0.130 CON>1 – SP Unit root -1.504 0.532 MOR – SP Unit root -1.461 0.553 MOR<1 – SP Unit root -1.251 0.651 MOR>1 – SP Unit root -1.304 0.627 DEP – SP Unit root 0.454 0.983 DEP<1 – SP Unit root 0.150 0.969 DEP>1 - SP Unit root 0.225 0.974 EURIBOR3M Unit root -1.479 0.544 Note: ***, denotes a rejection of the null hypothesis at a 10 percent significance level 5.2 Engle-Granger Augmented Dickey-Fuller (EG-ADF) test for cointegration
Following the results of the unit root test, the EG-ADF test for cointegration can be performed. The results for the first bank rate (the interest rate on new business loans to
non-financial corporations for Germany) are given in Table 5. The results for the other rates can be found in Table 6 in the appendix. Table 5: Results for the EG-ADF test for NFC Germany Interest rate Constant α Slope EURIBOR3M θ R2 Adj. R2 t-statistic NFC – GE 1.921* (0.023) 0.826* (0.020) 0.949 0.949 -3.075** Note: *,**,*** denote a rejection of the null hypothesis at a 1, 5, 10 percent significance level respectively The constant of 1.921 shows that this bank rates remains on average 1.921% higher than the EURIBOR. This can also be seen when looking at Figure 1 in the appendix. Moreover, the cointegration coefficient for this bank rate is 0.826, which indicates that the retail rate and the EURIBOR are cointegrated with a cointegration factor of 0.826. The t-statistic has a value of -3.075, which is lower than the critical value of -2.76 (see section 3.2.3). Therefore the null-hypothesis of no cointegration is rejected. This means that indeed this retail rate and the EURIBOR are cointegrating according to this EG-ADF test.
This result of a cointegrating relation between the retail rate and the EURIBOR can be found for most of the German retail rates, except for NFC<1, CON<1, MOR and MOR<1. If one looks at Figure 1 in the appendix it becomes clear that for example, for CON<1 it is indeed logical that this shows no cointegration relation, because CON<1 has a very atypical movement. It is the only rate that goes up over time.
For the Spanish bank rates however, a very interesting result is found. Namely, that almost none of the interest rates shows cointegration with the EURIBOR. Only NFC>1 and CON do. When looking at the Figure 2 in the appendix, one can see that the bank rates indeed follow the EURIBOR less smoothly than those of Germany. However, the fact that so little of the rates show cointegration with the EURIBOR is interesting. This could indicate that the Spanish banks indeed do not follow the EURIBOR closely, because of the lower level of competition in the banking sector. It could also mean that they maybe follow another rate better (for example the EONIA).
However, there is some critique towards the EG-ADF test. For example, as explained in section 4.2.3 this method is a two-step procedure. Therefore errors in the first estimation are automatically carried into the second estimation. This could lead to a type II error; when the null hypothesis of no cointegration is not rejected when in fact it is false. This type II error could be the case for some of the interest rates for Spain that show no cointegration. Therefore, this paper will now turn to the ECM to find out what short-term and long-term estimates are found, and to see if the results mentioned above are found in this model as well. 5.3 Single Equation Error Correction Model After doing the ADF test and the AG-ADF test, the single equation ECM is estimated using a regression. As discussed before, the regression then has to be rewritten slightly to put it back into the original model. Because this is done by hand, the standard errors for the final pass-through are not given in the tables below. The results for different loans and deposits are separately given and discussed in the following paragraphs. By separately analyzing these loans and deposits, it is possible to test for the presence of sectoral differences in the responsiveness of the interest rates to changes in the EURIBOR. In section 5.3.1 the results for interest rates on new business loans to non-financial corporations (NFC) will be given. This type of loan is used, because in the Euro area retail banks play a leading role in providing loans to non-financial corporations (ECB, 2009). Following, section 5.3.2 discusses interest rates on new loans to households for consumption (CON). In previous literature this rate shows to be more stable, and less affected by changes in the interbank rate than the other rates (Aristei, & Gallo, 2014). Therefore, it is interesting to test this rate to find if this is also the case for this dataset. Section 5.3.3 discusses the results for interest rates on new loans to households for house purchases (MOR), to see if the pass-through effect is different in the mortgage market. Lastly, the results for interest rates agreed on new deposits from private households (DEP) are discussed in section 5.3.4. This interest rate is chosen to see if there is indeed a difference in the interest rate pass-through between loans and deposits, as suggested by the literature (see section 2.2). The interest rates are also divided into short-term and long-term rates, to test if the pass-through is affected differently for short or long maturities. The usage of specifically these interest rates is
5.3.1 New business loans to non-financial corporations Firstly, the results for interest rates on new business loans to non-financial corporations (NFC) are summarized in Table 7. Table 7: Results ECM for NFC Interest rate Immediate pass-through α0 Final pass-through β Speed of adjustment θ R2 Adjusted R2 NFC - GE 0.946* (0.142) 1.006 -0.166** (0.07) 0.649 0.630 NFC<1 - GE 1.006* (0.141) 0.951 -0.267* (0.081) 0.655 0.636 NFC>1 - GE -0.175 (0.138) 1.356 -0.045 (0.035) 0.270 0.230 NFC - SP 0.845* (0.247) 1.313 -0.048 (0.044) 0.382 0.347 NFC<1 - SP 0.823* (0.242) 1.356 -0.045 (0.043) 0.406 0.373 NFC>1 - SP 1.012*** (0.589) 1.028 -0.145** (0.065) 0.183 0.137 Note: *,**,*** denote a rejection of the null hypothesis at a 1, 5, 10 percent significance level respectively Starting with the average NFC for Germany, the speed of adjustment (θ) has an estimate of -0.166. The fact that it is a negative number is in line with the hypothesis, namely that the bank interest rate grows towards the equilibrium with the market interest rate. If it is higher than the equilibrium, next period it will go down and vice versa. -0.166 Means that 16.6% of the bank rate’s divergence from its equilibrium with the market rate is corrected for in the following month. Very high numbers for the immediate and final pass-through are found, respectively 0.946 and 1.006. This means that the immediate pass-through is already almost complete, and the final pass-through is more than complete. A 1 percent increase of the EURIBOR leads to a 1.006 percent increase in the interest rate in the long-run.
This loan is also separated into one with maturities of shorter than a year (short-term loans), and one with maturities of longer than a year (long-term loans). Short-term loans have a higher immediate pass-through than the final pass-through. Moreover, the pass-through on interest rates with a maturity of longer than a year starts up slowly (with a low speed of adjustment and a negative immediate pass-through), but ends up more complete.
Comparing this to the Spanish interest rates we find some surprising results for the long-run. For the overall rate and the short-term rates, the pass-through for Germany is faster and the short-term pass-through is more complete. However, the final pass-through is more complete for Spain than for Germany. Though, this is accompanied by larger standard errors for the Spanish rates. For the rates of over one year, the opposite is found, namely a faster pass-through that is higher in the short-run (with a very large standard error), but a lower final pass-through for Spain. The R2 and adjusted R2 also differ greatly amongst the interest rates.
Comparing this to previous research, these results actually do not differ too much. For example, Sorensen and Werner (2006) did research on the Euro area before the crisis and they also found a higher final pass-through for short-term loans for Spain. What is different from what they found however is the speed of adjustment for these short-term loans, namely -0.046 and -0.925 for Germany and Spain respectively. This means that after the crisis this speed increased for Germany and decreased a lot for Spain (as I find -0.267 and -0.045 respectively).
Moreover, these more than complete final pass-through estimates for loans to non-financial companies can be explained by the presence of risky borrowers in this market segment. Banks in this segment increase their interest rates more than proportionally to compensate for the decline in the creditworthiness of borrowers (Andries & Billon, 2015). 5.3.2 New loans to households for consumption Next, the results for interest rates on new loans to households for consumption (CON) are given in Table 8.
Table 8: Results ECM for CON Interest rate Immediate Pass-through α0 Final Pass-through β Speed of adjustment θ R2 Adjusted R2 CON - GE 0.516 (0.356) 0.486 -0.251* (0.075) 0.149 0.102 CON<1 - GE -0.906** (0.433) -1.392 -0.171* (0.053) 0.193 0.148 CON>1 - GE 0.129 (0.238) 0.801 -0.261* (0.078) 0.167 0.120 CON - SP -0.882 (0.569) -0.005 -0.213* (0.064) 0.154 0.107 CON<1 - SP -0.768 (0.937) 0.832 -0.137* (0.047) 0.127 0.078 CON>1 - SP 0.051 (0.349) 0.685 -0.073 (0.051) 0.715 0.669 Note: *,**,*** denote a rejection of the null hypothesis at a 1, 5, 10 percent significance level respectively Immediately one result stands out; the rate for loans shorter than a year for Germany. This rate has a negative immediate and final pass through. If one looks at Figure 1 in the appendix one indeed sees that the rate moves upwards instead of downwards, therefore this result of the ECM is not surprising. For the average CON of Spain we also se a negative final pass-through, but it is of a smaller size.
For the average interest rates, and those with a maturity of over 1 year the hypothesis of a more complete and faster pass-through in Germany than in Spain is accepted. For these rates the speed of adjustment, the immediate and the final pass-through are all larger for Germany than for Spain. For the loans with a maturity of shorter than one year Spain also has a negative short-run pass-through but this is corrected for in the long-run. The R2 and adjusted R2 are relatively the same for all rates, except for the CON>1 for Spain. They are on average lower than what is found for all other rates for loans and deposits.
The results for the loans to households for consumption are very interesting if we compare it to previous research. For the average consumer rates previous literature from before the financial crisis found a higher speed of adjustment and final pass-through for Spain (Sorensen & Werner, 2006; Bernhofer & van Treeck, 2013). The fact that the recent data that is used in this research results in the opposite, may indicate that the higher level of competition in the German banking sector compared to the lower one of Spain indeed led to a relative worsening of the Spanish pass-through effect when looking at this specific interest rate. 5.3.3 New loans to households for house purchases Following, the results for interest rates on new loans to households for house purchases (MOR) are given and interpreted. Table 9 gives an overview of the estimates of the ECM. Table 9: Results ECM for MOR Interest rate Immediate Pass-through α0 Final Pass-through β Speed of adjustment θ R2 Adjusted R2 MOR - GE 0.094 (0.109) 0.770 -0.003 (0.016) 0.328 0.291 MOR<1 - GE 0.303 (0.144) 0.858 -0.212* 0.071 0.583 0.560 MOR>1 - GE 0.084 (0.111) 0.667 -0.003 (0.017) 0.303 0.265 MOR - SP 0.129 (0.099) 1.138 -0.029 (0.025) 0.714 0.699 MOR<1 - SP 0.143 (0.105) 1.379 -0.029 (0.026) 0.699 0.681 MOR>1 - SP 0.277*** (0.150) 1.230 -0.074** (0.037) 0.425 0.393 Note: *,**,*** denote a rejection of the null hypothesis at a 1, 5, 10 percent significance level respectively
Looking at these results a low speed of adjustment for both countries is found, which was also found in previous literature about this rate in these two countries (Bernhofer & van Treeck, 2013). More than complete final pass-through estimates are seen for Spain and less than complete for Germany. The hypothesis that Spain would have a less complete and slower pass-through than Germany is completely rejected in this section. For almost all rates and maturities a slower pass through for Germany is seen, except for short-term loans. For this rate also a larger immediate pass-through than for Spain is found. For the immediate and final pass-through for all the other maturities Spain outperforms Germany. The R2 values found are large, and the standard errors are low compared to other types of loans and deposits.
Next to rejecting the hypothesis for the speed and completeness for these two countries, if these results are compared to previous research from before the crisis this also rejects the theory that a lower level of competition leads to a lower pass-through. For example, Bernhofer and van Treeck (2013 ) find a faster and more complete through for Germany, whereas these results show a faster and more complete pass-through for Spain.
Furthermore, it is found that indeed the pass-through becomes more complete over time, which is in line with the literature (Andries & Billon, 2015). This is seen by the fact that for all interest rates the long-run pass-through is more complete than the short-run pass-through. This is explained by information asymmetry in the short-run that disappears over time (Cottarelli & Kourelis, 1994).
5.3.4 New deposits from private households
Lastly, the results for interest rates agreed on new deposits from private households (DEP) are summarized in Table 10.
Table 10: Results ECM for DEP Interest rate Immediate Pass-through α0 Final Pass-through β Speed of adjustment θ R2 Adjusted R2 DEP - GE 0.359* (0.112) 0.623 -0.167* (0.063) 0.645 0.626 DEP<1 - GE 0.298** (0.118) 0.618 -0.275* (0.071) 0.678 0.660 DEP>1 - GE -0.018 (0.203) 0.956 -0.091*** (0.047) 0.253 0.211 DEP - SP 0.417** (0.162) 1.000 -0.013 (0.039) 0.417 0.385 DEP<1 - SP 0.430** (0.193) 1.036 -0.028 (0.028) 0.335 0.298 DEP>1 - SP 0.444*** (0.225) 1.658 -0.038 (0.032) 0.263 0.222 Note: *,**,*** denote a rejection of the null hypothesis at a 1, 5, 10 percent significance level respectively Again higher immediate and final pass-through estimates are found for Spain than for Germany. This rejects the hypothesis that the completeness would be larger for Germany than for Spain in this time frame. The Spanish rates show a complete or more than complete final pass-through, whereas for Germany only the long-term rate shows a complete final pass-through. However, for all rates the speed of adjustment is higher for Germany. Therefore, for these rates the hypothesis of a faster pass-through for Germany is accepted. The R2 and adjusted R2 are quite high and mostly larger for Germany than for Spain. For these interest rates it is found that the final pass-through for long-term rates is higher than for those with a short maturity. This is in correspondence with the results of Bondt (2005), and can be explained by the existence of crucial switching costs in the market of demand and savings deposits. Moreover, literature also shows on average lower final pass-through for interest
De Bondt (2005) explains this by the fact that the elasticity of demand for these loans is bigger than the elasticity of demand for deposits. This is also completely seen in these results for Germany, and partly for Spain. Only long-term Spanish deposits show a higher final pass-through than Spanish loans to non-financial corporations.
Furthermore, literature about the pass-through before the crisis found a faster pass-through for Spain, whereas this research finds a higher speed of adjustment for Germany (Sorensen & Werner, 2006). The level of completeness found in that research was also larger for Spain, but the more resent data used in this paper shows that this difference grew even larger. Therefore again it can be stated that the hypothesis of a faster speed for Germany can be accepted, whereas the hypothesis of a more complete pass-through is rejected. The possible explanation for this higher speed can be the higher level of competition in the banking sector.
6. Conclusion
In this paper the interest rate pass-through effect for Germany and Spain was estimated and discussed. The aim of this was to see what the effect of competition in de banking sector in a country is on the overall pass-through. To do so, several retail banking interest rates (on loans to non-financial corporations, loans to households for consumption, loans to households for house purchase and deposits from households) were used. Additionally, the 3-month EURIBOR was used as a proxy for the market rate. Moreover, monthly data from September 2008 until October 2016 was used. A single equation Error Correction Model was regressed to estimate the immediate pass-through, the final pass-through and the speed of adjustment towards the equilibrium between the retail rate and the market rate. The results of these regressions led to several findings.
Firstly, the speed of adjustment is found to be higher for Germany than for Spain for almost all rates. This means that the hypothesis for a faster pass-through in a country with more competition in the banking sector can be accepted for this dataset. Compared to literature from before the crisis it even showed that speed of adjustment of the interest rates of NFC<1, CONS and DEP were first larger for Spain but are now found to be higher for Germany.
Secondly, these findings are less clear when looking at the completeness of the pass-through. Significant differences between the levels of completeness are found
between the different types of loans and deposits. In the final completeness higher values for Spain than for Germany are found (except for NFC>1, CON, CON>1). Old literature with data from before the crisis shows that this was already seen then. Only CON and CON>1 became more complete for Germany than for Spain since the crisis. Therefore the hypothesis of a more complete pass-through for countries with a higher level of competition in the banking sector is not accepted for this dataset.
Overall, the CON rates behaved the most as expected, whereas MOR behaved completely the opposite. Moreover, despite the fact that there are several differences among the short-term and long-term interest rates in terms of speed and completeness, almost all interest rates respond to changes in the market rate with a lag. This means that on average it is found that the immediate pass-through is less complete than the final pass-through, which coincides with the literature on this subject. Lastly, an interesting result is that some interest rates show a negative immediate or even final pass-through. However, there are arguments as to why these results could have happened that rely more on the power of the specific model that is used. For example, as seen in the EG-ADF test there was no cointegration relation found for a lot of Spanish interest rates and the EURIBOR. This could indicate that these long-term estimates are not completely right for these rates. Notwithstanding, maybe the EURIBOR is not the best proxy for the market rate set by the ECB. Table 12 in the appendix shows the results for the ECM if the EONIA rate is used as a proxy. The difference between the EONIA and the EURIBOR is that the EONIA measures the interest rate in the overnight interbank market. It therefore does not include term loans, whilst the EURIBOR does (Mishkin et al., 2013). As one can see, the EONIA rate give different results than the EURIBOR, when used in the ECM. It inflates most of the estimates even further in a positive or negative direction.
Some researchers try to tackle this problem by matching different interest rates to different market rate proxies (Sander & Kleimeier, 2004). However, there is also critique to this method, because matching certain rates to different market rates means that one will influence the results of the ECM beforehand, which will probably result in a bias. Therefore in this paper it was chosen to stick to one proxy, as this would be more reliable. Notwithstanding, it is still important to be informed about the fact that this ECM gives such different results if one uses another market proxy or retail bank interest rate.
It definitely raises the question if this model is the best model to test the pass-through effect with, or if it could be improved to diminish this certain ‘randomness’. All together theses findings lead to several recommendations for future research. Firstly, there should be more research into the pass-through of Spain and compare it to other individual countries with lower banking competition. This way the heterogeneity between countries in the Euro Area could be explained further. Secondly, this ECM does not provide the opportunity to test what the exact explanation is for the estimates of the pass-through. To test the real effect of the banking competition the Herfindahl-Index or the market share of the largest five banks in a country (see section 3) could be added to the model as variables. To do so however, many more countries and HI observations need to be added and this was not possible within the scope of this bachelor thesis. Moreover, the previous paragraph shows the weakness of this Error Correction Model, since it is very dependent on which rates are used. Future research should therefore focus on the reliability of this model itself, and how to improve it. All of these recommendations would help to further develop the research of the subject of the interest rate pass-through, which remains important especially in recent times with the new phenomena of negative interest rates.
References
Abbassi, P., & Linzert, T. (2012). The effectiveness of monetary policy in steering money market rates during the financial crisis. Journal of Macroeconomics, 34(4), 945-954.
Andries, N., & Billon, S. (2015). Retail bank interest rate pass-through in the euro area: An empirical survey. Economic Systems, 40(1), 170-194.
Aristei, D., & Gallo, M. (2014). Interest rate pass-through in the Euro area during the financial crisis: A multivariate regime-switching approach. Journal of Policy Modeling, 36(2), 273-295.
Bernhofer, D., & van Treeck, T. (2013). New evidence of heterogeneous bank interest rate pass-through in the euro area. Economic Modelling, 35, 418-429.
Ciccarelli, M., Maddaloni, A., Peydró, J.-L. (2010). Trusting the bankers: A new look at the credit channel of monetary policy. Working Paper Series 1228, European Central Bank.
Cottarelli, C., & Kourelis, A. (1994). Financial structure, bank lending rates, and the transmission mechanism of monetary policy. Staff Papers, 41(4), 587-623.
De Bondt, G. (2002). Retail bank interest rate pass-through: new evidence at the euro area level. (working paper)
De Bondt, G. J. (2005). Interest Rate Pass-Through: Empirical Results for the Euro Area. German Economic Review, 6(1), 37-78.
De Graeve, F., De Jonghe, O., & Vander Vennet, R. (2007). Competition, transmission and bank pricing policies: Evidence from Belgian loan and deposit markets. Journal of Banking & Finance, 31(1), 259-278.
Engle, R. F., & Granger, C. W. (1987). Co-integration and error correction: representation, estimation, and testing. Econometrica: journal of the Econometric Society, 251-276.
European Central Bank. (2003, October). Manual on MFI interest rate statistics.
European Central Bank. (2009, August). Recent developments in the retail bank interest rate pass-through in the Euro area. ECB Monthly bulletin, 93-105.
European Central Bank. (2014, October). ECB Banking structures report.
European Central Bank. (2016, March). Key ECB interest rates. Retrieved from
https://www.ecb.europa.eu/stats/monetary/rates/html/index.en.html
Gambacorta, L., & Iannotti, S. (2007). Are there asymmetries in the response of bank interest rates to monetary shocks?. Applied Economics, 39(19), 2503-2517.
Hamilton, J. D. (1994). Time series analysis (Vol. 2). Princeton: Princeton university press.
Hristov, N., Hülsewig, O., & Wollmershäuser, T. (2014). The interest rate pass-through in the Euro area during the global financial crisis. Journal of Banking & Finance, 48, 104-119.
Johansen, S. (1991). Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models. Econometrica: Journal of the Econometric Society, 1551-1580.
Karagiannis, S., Panagopoulos, Y., & Vlamis, P. (2010). Interest rate pass-through in Europe and the US: Monetary policy after the financial crisis. Journal of Policy Modeling, 32(3), 323-338.
Kleimeier, S., & Sander, H. (2006). Expected versus unexpected monetary policy impulses and interest rate pass-through in euro-zone retail banking markets. Journal of Banking & Finance, 30(7), 1839-1870.
Kok Sorensen, C., & Werner, T. (2006). Bank interest rate pass-through in the euro area: a cross country comparison. (working paper)
Lowe, P., Rohling, T. (1992). Loan rate stickiness: theory and evidence. Research Discussion Paper 9206. Reserve Bank of Australia.
Mishkin, F.S., Matthews, K., & Giuliodori, M. (2013). The Economics of Money, Banking & Financial Markets. Pearson Education.
Rocha, M. D. (2012). Interest rate pass-through in Portugal: Interactions, asymmetries and heterogeneities. Journal of Policy Modeling, 34(1), 64-80.
Sander, H., & Kleimeier, S. (2004). Convergence in euro-zone 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.
Scholnick, B. (1996). Asymmetric adjustment of commercial bank interest rates: evidence from Malaysia and Singapore. Journal of International Money and Finance, 15(3), 485-496.
Schwert, G. W. (2002). Tests for unit roots: A Monte Carlo investigation. Journal of Business & Economic Statistics, 20(1), 5-17.
Statistical Data Warehouse. (2016, December). Structural financial indicators. Retrieved from http://sdw.ecb.europa.eu/reports.do?node=1000002869
(retrieved: 13-12-Stock, J. H., & Watson, M. W. (2015). Introduction to econometrics. Boston: Pearson/Addison Wesley.
Appendix Figure 1: Overview of the retail rates and the EURIBOR (Germany) Figure 2: Overview of the retail rates and the EURIBOR (Spain) -1 0 1 2 3 4 5 6 7 8 9 se p. -0 8 ja n. -0 9 me i-0 9 se p. -0 9 ja n. -1 0 me i-1 0 se p. -1 0 ja n. -1 1 me i-1 1 se p. -1 1 ja n. -1 2 me i-1 2 se p. -1 2 ja n. -1 3 me i-1 3 se p. -1 3 ja n. -1 4 me i-1 4 se p. -1 4 ja n. -1 5 me i-1 5 se p. -1 5 ja n. -1 6 me i-1 6 se p. -1 6 Interest rate (in %) Date NFCGE NFCS1GE NFCB1GE CONGE CONS1GE CONB1GE MORGE MORS1GE MORB1GE DEPGE DEPS1GE DEPB1GE EURIBOR3M 2 4 6 8 10 12 14 Interest rate (in %) NFCSP NFCS1SP NFCB1SP CONSP CONS1SP CONB1SP MORSP MORS1SP MORB1SP DEPSP DEPS1SP DEPB1SP
Table 6: Engle-Granger Augmented Dickey-Fuller (EG-ADF) test for cointegration Rate CONSTANT α Slope EURIBOR3M θ R2 Adj. R2 t-statistic NFC – GE 1.921* (0.023) 0.826* (0.020) 0.949 0.949 -3.075** NFC<1 – GE 1.789* (0.019) 0.831* (0.016) 0.965 0.964 -3.976** NFC>1 – GE 2.619* (0.067) 0.865* (0.056) 0.716 0.713 -2.246 CON – GE 6.437* (0.056) 0.175* (0.047) 0.127 0.118 -3.230** CON<1 – GE 4.793* (0.096) -0.135** (0.080) 0.029 0.019 -2.595 CON>1 – GE 6.641* (0.062) 0.270* (0.052) 0.222 0.214 -2.957** MOR – GE 2.530* (0.063) 0.819* (0.052) 0.719 0.716 -2.201 MOR<1 – GE 2.432* (0.025) 0.853* (0.021) 0.947 0.950 -4.261** MOR>1 – GE 2.558* (0.071) 0.810* (0.060) 0.659 0.655 -2.116 DEP – GE 0.644* (0.020) 0.767* (0.017) 0.956 0.955 -3.225** DEP<1 – GE 0.484* (0.020) 0.770* (0.016) 0.958 0.958 -3.518** DEP>1 - GE 1.278* (0.049) 0.832* (0.041) 0.812 0.810 -3.211** NFC – SP 2.834* (0.061) 0.535* (0.051) 0.533 0.528 -2.063 NFC<1 – SP 2.808* (0.060) 0.539* (0.050) 0.543 0.539 -2.080
NFC>1 – SP 3.231* (0.102) 0.483* (0.085) 0.252 0.244 -3.111** CON – SP 8.318* (0.107) 0.495* (0.089) 0.243 0.235 -2.923** CON<1 – SP 5.834* (0.224) 1.441* (0.187) 0.383 0.377 -2.427 CON>1 – SP 9.20* (0.075) 0.094 (0.062) 0.023 0.0131 -2.293 MOR – SP 2.464* (0.046) 0.750* (0.038) 0.800 0.799 -1.551 MOR<1 – SP 2.260* (0.046) 0.805* (0.038) 0.820 0.829 -1.617 MOR>1 – SP 2.923* (0.046) 0.788* (0.038) 0.817 0.815 -2.070 DEP – SP 1.210** (0.075) 0.958** (0.062) 0.710 0.707 -1.644 DEP<1 – SP 1.117* (0.076) 0.978* (0.063) 0.715 0.711 -1.799 DEP>1 - SP 1.307* (0.077) 0.933* (0.064) 0.687 0.683 -2.367 Note: *,**,*** denote a rejection of the null hypothesis at a 1, 5, 10 percent significance level respectively Table 11: results ECM using EONIA for Germany Interest rate Immediate Pass-through α0 Final Pass-through β Speed of adjustment θ R2 Adjusted R2 NFC 0.268* (0.138) 1.552 -0.181* (0.058) 0.306 0.265 NFC<1 0.278*** (0.144) 1.449 -0.247* (0.066) 0.342 0.304 NFC>1 0.367*** 1.976 -0.041 0.170 0.121
CON 0.335 (0.376) 0.759 -0.249* (0.077) 0.131 0.080 CON<1 -0.676 (0.386) -2.044 -0.136** (0.057) 0.151 0.102 CON>1 0.518*** (0.365) 1.227 -0.256* (0.080) 0.142 0.092 MOR -0.046 (0.094) 1.750 -0.008 (0.019) 0.200 0.152 MOR<1 0.106 (0.156) 1.326 -0.258* (0.062) 0.312 0.272 MOR>1 -0.039 (0.094) 1.167 -0.006 (0.017) 0.258 0.215 DEP 0.115 (0.118) 1.076 -0.119** (0.050) 0.230 0.185 DEP<1 0.148 (0.122) 0.983 -0.180* (0.059) 0.272 0.230 DEP>1 0.037 (0.186) 1.529 -0.104** (0.043) 0.086 0.032 Note: *,**,*** denote a rejection of the null hypothesis at a 1, 5, 10 percent significance level respectively Table 12: results ECM using EONIA for Spain Interest rate Immediate Pass-through α0 Final Pass-through β Speed of adjustment θ R2 Adjusted R2 NFC 0.661** (0.266) 1.632 -0.057 (0.045) 0.195 0.148 NFC<1 0.610** (0.261) 1.625 -0.056 (0.045) 0.195 0.148 NFC>1 0.823 (0.644) 1.568 -0.155** (0.068) 0.184 0.137 CON -0.944 (0.612) 0.256 -0.207** (0.068) 0.150 0.100
CON<1 -1.518 (1.009) 1.889 -0.144* (0.046) 0.143 0.093 CON>1 -0.208 (0.378) 1.320 -0.075 (0.051) 0.128 0.077 MOR 0.132 (0.116) 1.255 -0.051 (0.026) 0.396 0.361 MOR<1 0.151 (0.122) 1.471 -0.051*** (0.027) 0.365 0.328 MOR>1 0.433** (0.169) 1.637 -0.102* (0.038) 0.205 0.159 DEP 0.158 (0.154) 2.500 -0.020 (0.023) 0.156 0.107 DEP<1 0.010 (0.187) 2.361 -0.036 (0.026) 0.122 0.071 DEP>1 0.568* (0.199) 2.367 -0.030 (0.032) 0.170 0.122 Note: *,**,*** denote a rejection of the null hypothesis at a 1, 5, 10 percent significance level respectively
