An empirical study on the interest rate pass-through effect in the euro area

(1)

Faculty of Economics & Business

AN EMPIRICAL STUDY ON THE INTEREST

RATE PASS-THROUGH EFFECT IN THE EURO

AREA

BY

MIKEY STAATS

10336613

Thesis Supervisor: Lin Zhao

University of Amsterdam

May-June 2015

(2)

The interest rate pass-through is one of the most important channels in the Euro area through, which monetary policy can influence the real economy. In the years up to the global financial crisis, the interest rate pass-through was complete and seemed to produce the desired effects. However, during the years following the financial crisis several studies have shown that the interest rate pass-through is far from perfect. Short and long-term interest rates seem to respond in a slow and sticky manner to market interest rate changes, while the final pass-through is far from complete. This paper tries to identify whether this channel is properly working for the time period January 2009 till March 2015 by using an error correction model. Empirical results from this study show that indeed the pass-through effect is distorted and that both short and long-term interest rates respond slowly in response to market interest rate changes. Furthermore it is found that the final pass-through effect is far from complete.

1. Introduction

The year 2008 marked the beginning of a global financial crisis. Since then the United States managed to take matters into own hands and turned the tide. The Euro area however, is still experiencing turmoil on the financial markets and low to decreasing economic growth in response to the aftermath of the global financial crisis.

Furthermore banks and other financial institution have significantly tightened their credit standards, thereby drastically lowering lending activities. In order to stimulate nominal spending and investing on a national level, the European Central Bank (ECB) began by lowering the official policy rates drastically at the end of 2008 and

continues to do so. By lowering the official policy rates the ECB tries to induce banks and other financial institutions to increase their lending activities and thereby

ultimately increasing national economic growth. The effectiveness of lowering official policy rates on the real economy is often determined by the their effect on other interest rates. In particular, the adjustment of retail bank interest rates in response to changes in the official policy rates is of utmost importance to stimulate the national economies in the Euro area. This phenomenon is most common referred to as the interest rate pass-through effect. In general, the effectiveness in terms of speed and completeness in the pass-through effect is determined by several structural and cyclical factors. Therefore, the retail bank interests rates typically adjust to changes in the official policy and market rates with a lag.

(4)

According to the ECB (2009) the interest rate pass-through seemed to have worked properly during the years before the financial crisis, however recent data suggests a less than smooth through to retail interest rates. Since the interest rate pass-through effect is of fundamental importance in stimulating the national economies of the Euro area, it is of utmost importance to determine whether this particular channel of Monetary Transmission Mechanisms (MTMs) is working properly. The following paper will give an extensive empirical study on the interest rate pass-through effect in the Euro area for the period 2009-2015 to address this question.

This paper commences with the theoretical background necessary to fully comprehend the empirical study as well as an overview of previous literature and research in Section 2. In this section a reflection is given of the contemporary findings on the interest rate pass-through effect as well as recent developments.

Section 3 addresses the specific research method used for the empirical study as well as a description of the data. Section 4 provides the results of the empirical study and Section 5 discusses the results. Finally, Section 6 concludes.

2. Literature Review

In order to fully comprehend the interest rate pass-through effect an extensive description must be given of MTMs, since the interest rate pass-through effect is one of many channels through which monetary policy influences the real economy. Therefore this section starts with the basic theory necessary to understand the tools of monetary policy, their corresponding channels and the interest rate pass-through effect in particular. Finally, it concludes with an overview of the extensive number of studies conduction in the past three decades on the interest rate pass-through effect.

Monetary Transmission Mechanisms

The ultimate goal of the ECB is to main its price stability mandate of achieving inflation rates below, but close to, 2% over the medium term. This mandate can be achieved by using so-called tools of monetary policy. There are several tools of monetary policy, which can be divided into two different categories namely: conventional and unconventional.

(5)

Traditionally, the ECB uses main refinancing operations, deposit/lending facilities and minimum reserve requirements, but recently it started to implement

unconventional tools, including quantitative easing and forward guidance. Decisions on monetary policy are transmitted through the economy in different ways, which ultimately have a direct and indirect effect on the movement of prices of goods and services. The combination of different channels through which these prices are affected is commonly referred to as Monetary Transmission Mechanisms (MTMs). The literature distinguishes between three different channels, namely traditional interest rate channels, other asset price channels and credit view channels (Mishkin, 1996, pp. 2-15). Although the specific chain of events may vary between these channels, all share the same starting point, namely the official ECB interest rates or the monetary base. Generally, the MTM has two broad stages. During the first stage, changes in the official policy rates or in the monetary base lead to changes in general financial market conditions, including asset prices, exchange rates, credit and liquidity conditions and market interest rates. The second stage is characterized by changes in nominal spending and investments by firms and households. According to Bernanke and Gertler (1995) changes in nominal spending and investments may have a

significant impact on the real sector of the economy in the short-run. The eventual impact on the real economy will be ultimately determined by its flexibility and degree of nominal rigidities in wages and prices. In the long run however, these nominal changes only affect the general price level, which is part of the ECB’s mandate (ECB, 2000, pp. 43-44).

The interest rate pass-through effect

According to the ECB (2009), out of all MTMs the interest rate pass-through effect is of fundamental importance in the Euro area due to its overall bank-based financial system. The interest rate pass-through effect is defined as the adjustment of retail interest rates on loans and deposits in response to changes in the official policy rates. It is particularly important for countries within the Euro area, which have bank-based financial systems and where banks are of relevant importance when it comes to providing financing to and collecting savings from the non-financial sectors of the economy. The main idea behind the interest pass-through effect is as follows.

Changes in the official policy rates cause movements in several corresponding market rates with different maturities through the yield curve.

(6)

Subsequently, the changes in the market interest rate pass through to the

corresponding retail rates. These retail interest rates will affect investments and consumption by households and companies, which ultimately determines general price levels in the Euro area. Therefore the interest rate pass-through effect contributes to the ECB’s mandate of achieving price stability in the Euro area (Gigineishvili, 2011, p.5).

Determinants of the interest rate pass-through effect

Historically, retail interest rates have been moving closely in line with their

corresponding market interest rates. However, it is also clear in the short run that in some cases the retail deposit and lending rates have been reacting rather slowly to changes in market interest rates. Also when comparing the long-term reaction of retail interest rates to market interest rates a less than perfect co-movement is seen (ECB, 2009, pp.84-85). According to Illes and Lombardi (2013) this so-called spread between retail and market interest rates can be seen as a rough indicator of the effectiveness of the interest rate pass-through effect. Although it is far too early to make any conclusions about the effect of the interest rate pass-through, it is suffice to say that it is far from perfect in the Euro area. In general how quickly and extensive changes in the official policy rates are passed through to retail bank interest rates is determined by several structural and cyclical factors. Cottarelli and Kourelis (1994) together provide a theoretical point of view to explain this observed lag in banks’ interest rate-setting behavior. Typically, the price-setting behavior of banks is

considered from an oligopolistic competition model point of view. In the oligopolistic competition model, the assumption is made that banks act as price-setters while taking into account the demand elasticity for loans and deposits. Simultaneously banks are assumed to be acting as price-takers in the interbank market for attracting additional funding or storing excessive liquidity. Given these assumptions and a certain degree of market power, bank lending and deposit rates are normally considered to provide an upper and lower bound on the market interest rate. However, this may not always be the case in the Euro area. In particular, banks’ price-setting behavior is affected by the economy‘s financial structure, such as the degree of competition within the banking system, and between banks and other financial intermediaries. Also to be considered are constraints on capital movements, the ownership structure of financial intermediaries and banks’ market power (Contarelli & Kourelis, 1994, p.591).

(7)

Besides the economy’s financial structure individual differences among banks also cause differences in price-setting behavior. Factors related to the bank’s cost of intermediation, among which interest rate risk, credit risk, risk aversion, operating costs, liquidity and product diversification are also to be considered. Furthermore adjustment costs, non-profit maximizing behavior, asset quality, customer

relationships, uncertainty about future money market rates and refinancing conditions have a considerable impact on the effectiveness of the interest rate pass-through (Weth, 2002, pp. 2-6). Finally, other structural characteristics, like an economy‘s regulatory environment, exchange rate regime and inflation rate are likely to influence price-setting behavior as well (Saborowski & Weber, 2013, pp. 8-11). In addition to all these structural and cyclical factors, individual differences among the countries in the Euro area are also of great importance. The Euro area has a constantly changing composition, therefore individual differences in the factors mentioned earlier are increasingly becoming important in determining the effectiveness of monetary policy and to ensure the functioning of the interest rate pass-through effect.

Previous conducted studies

In the past three decades there have been quite an extensive number of studies examining the effect and characteristics of the interest rate pass-through in the Euro area. In most studies particular attention has been placed on whether there is a heterogeneous pass-through effect both in terms of the speed and degree of adjustment. The econometric methods used and the scope of research has differed widely among the various studies. There are for example studies that focus on individual countries or on the Euro area as a whole. Furthermore most of the econometric models used in previous research centered on single-equation error-correction models (ECM) and standard vector autoregression (VAR), but more recent studies have used marginal cost pricing models, a combination of panel unit root and cointegration tests and autoregressive distributed lag models (ADL)1. Previous research also differs from each other with respect to data sources, time interval covered and in the selection of exogenous variables.

1_{See Sørensen & Werner (2006); de Bondt (2002); Liu, Margaritis & Tourani-Rad (2007); Bernanke &}

(8)

Most of the earlier studies made extensive use of data from the International

Monetary Fund or from the Federal Reserve, whereas most recent studies use datasets on retail interest rates from the ECB or national central banks in the Euro area. The time period covered by previous studies ranges from the early 1980s to the present day, although most studies tend to focus on a range of five to ten years within this ragen. Finally, with regards to the exogenous variables, the majority of previous studies used money market interest rates as exogenous variable against which to measure the pass-through effect to retail interest rates. More recent studies however, use money market interest rates of comparable maturities to better reflect interest-rate setting behaviour by banks. Also some studies deviate from solely using market interest rates movements as means of explaining retail interest rate changes and use additional explanatory variables such as measures for bank competition, credit risk, bank balance sheet data, the extent of money market development, financial structure or the regulatory system2_{. Despite the large diversity of approaches to explain the}

interest rate pass-through effect, there seems to be some consensus on the degree and speed of the pass-through effect. Whereas earlier studies assumed an immediate and complete pass-through of changes in official policy rates to retail bank rates3, more recent studies conclude that the pass-through might be incomplete and adjustment speed relatively slow in the short-run. The pass-through and adjustment speed also differs considerably among individual countries in the Euro area as well as across financial institutions and banking products. The majority of studies also suggest that retail interest rates on savings, household loans and overnight deposits adjust

relatively slow, while the rates on time deposits and loans to enterprises adjust relatively quickly. With regards to the pass-through effect in the long run, there is no wide consensus and evidence is scattered. However, recent evidence seems to indicate a less than complete pass-through in the long run. Additionally, it is widely

recognized that there has been a significant structural change in the behavior of banks in terms of pricing-setting behavior, as the interest rate pass-through has become less complete after the crisis. Finally, with regards to the determinants of the interest pass-through effect there seems to be a lesser degree of consensus.

2_{See Hristov, Hülsewig & Wollmershäuser (2014); Angeloni, et al. (2003); others mentioned earlier} 3_{See Bernanke & Gertler (1995); Kashhyap & Stein (2000)}

(9)

Taking all these findings into account, this paper will extend the existing literature and will contribute to the ongoing debate on the speed and completeness of the interest rate pass-through effect by focusing on more recent data and using harmonized data collected by the ECB.

3. Methodology

As mentioned earlier, previous studies used an extensive and non-exhaustive variety of econometric models to measure the interest rate pass-through effect. Methods range from ECMs to VARs, each using different types of data. Since this paper will focus on the speed and completeness of the interest rate pass-through effect after the global financial crisis, the most important equation to be used is the single equation error-correction model (Sørensen & Werner, 2006, p. 21).

Description of the Data

The data used in this study is completely based on harmonised monthly monetary financial institutions (MFI) interest rate statistics. These statistics are retrieved from the Eurosystem of Central Banks (ESCB) and the data ranges from January 2009 to March 2015. The harmonised MFI data for interest rates is chosen to circumvent the issue of different classifications and definitions in interest rate categories among the countries within the Euro area. Previous studies on the interest rate pass-through effect in the Euro area have used non-harmonised national retail interest rates (NRIR), which can, due to the data issues mentioned earlier, cause heterogeneity among the test results by itself. Therefore by using harmonised MFIs interest rates for the whole Euro area instead of NRIRs, the bias present in NRIRs caused by different

classifications and definitions in interest rate categories will be limited and negligible. Furthermore the timeframe considered in this paper (2009-2015) is chosen for very specific reasons. Namely, in August 2007 deterioration in the value of US subprime mortgages triggered turbulence on the financial market that quickly spilled over to other segments of the financial market. The turbulence ensued and developed into a global financial crisis resulting in massive losses and asset write-downs by US as well as Euro area based banks (Tong & Wei, 2008, pp. 3-6). These losses and write-downs in turn resulted in significant pressures on liquidity and solvency ratios, causing loss of confidence in and among the banking sector, which worsened matters even further.

(10)

As a result of the lost confidence the functioning of the money market in the Euro area was heavily disrupted and caused a persistent and large widening of the interest rate spreads, which potentially impaired the interest rate pass-through effect in the Euro area (ECB, 2009, p. 99). Therefore to reduce any distortions in measuring the interest rate pass-through effect, data from right after the eruption of the global financial crisis is avoided. Additionally, February 2015 marked the launch of a new quantitative easing program initiated by the ECB. The new program entails a monthly purchase of assets for an amount up to €60 billion, through which in addition to earlier programs, bonds issued by Euro area central governments, agencies and European institutions will be bought. Asset purchases are to be carried out until September 2016 to ultimately fulfil the ECB’s price stability mandate (ECB, 2015). However to avoid any distortion caused in the data from the initiation of this new program, data after this date will be mostly omitted for the same reasons mentioned earlier. Table 1.1 exhibits general statistics for each variable used in the empirical study.

Table 1.1 # Obser. Mean Median Min Max

Short-term loans for house purchase (STBR1) 75 3.708 3.770 2.700 5.220

Short-term loans for consumer credit (STBR2) 75 5.764 5.440 4.840 8.180

Short-term loans to non-financial institutions (STBR3) 75 3.712 3.650 3.030 5.100

Overdrafts non-fin institutions (STBR4) 58 4.016 4.020 3.330 4.540

Overdrafts Households (STBR5) 58 7.901 7.985 7.070 8.470

Overnight Deposits non-fin institutions (STBR6) 75 0.483 0.450 0.210 1.260

Overnight Deposits Households (STBR7) 75 0.420 0.430 3.030 1.020

Short-term time deposits non-fin institutions (STBR8) 75 1.002 1.010 0.310 2.250

Short-term time deposits households (STBR9) 75 2.094 2.160 0.910 3.280

Savings Deposits (STBR10) 75 1.508 1.540 0.800 2.960

EONIA (STMR1) 75 0.388 0.340 0.050 1.810

EURIBOR three-months (STMR2) 75 0.711 0.680 0.030 2.460

EURIBOR six-months (STMR3) 75 0.905 0.970 0.097 2.539

Long-term loans for house purchase (LTBR1) 75 3.744 3.810 3.090 4.930

Long-term loans for consumer credit (LTBR2) 75 6.901 6.920 6.030 7.730

Long-term loans to non-fin institutions (LTBR3) 72 3.449 3.370 2.960 4.890 Long-term time deposits non-fin institutions (LTBR4) 75 2.208 2.310 0.920 3.870

Long-term time deposits Households (LTBR5) 75 2.357 2.440 1.050 3.700

Three-year swap rate (LTMR1) 75 1.115 1.000 0.460 2.260

(11)

Selection of Retail Interest Rates

To give a proper reflection of the interest rate pass-through effect a broad range of retail interest rates is chosen. Both loan as well as deposit rates are included to reflect the broad spectrum. Furthermore there will be a distinction between short-term and long-term interest rates to take differences in price-setting behavior into account for different maturities. Finally, there is also distinction between households and non-financial corporations to reflect differences in price-setting behavior resulting from different risk profiles, loan/deposit amounts and other factors. In practical terms, for the deposits the following retail interest rates are chosen: overnight, savings and short-term time deposits and long-term time deposits4. For the loans the following retail interest rates are chosen: short and long-term loans for house purchases, short and long-term loans for consumer credit and short and long-term loans to non-financial corporations5_{. Using the data for these specific retail interest rates with}

regard to the timeframe provides a total sample of 75 observations6_. Selection of corresponding Market Interest Rates

According to Sørensen and Werner (2006), to select the relevant market interest rates for individual retail interest rates a so-called cost-of-funds or opportunity cost

approach needs to be taken. Similarly as de Bondt (2002) calls it, a marginal cost pricing approach can be used. The starting point of both approaches is the assumption that retail interest rates are set in accordance with their marginal cost, which is

approximated by market interest rates with comparable maturities. Therefore for deposit rates the corresponding market rate can be seen as the cost-of-funds, which is taken into consideration by banks when determining their retail interest rates.

Similarly for lending rates the corresponding market interest rates represents the opportunity cost. Ultimately the aim of this approach is to combine market and retail interest rates that are highly related.

4_{Short-term time deposits are defined as having maturity up to one year, while long-term time deposits}

are defined as having a maturity of over one year. Furthermore savings are defined as deposit accounts redeemable at notice.

5_{Short-term loans are defined as having maturity up to one year, while long-term loans are defined as}

having a maturity of over one year.

6_{There are only 58 observations for the overdrafts for households and non-financial corporations.}

(12)

As proposed by the ECB (2009) the following market interest rates are used: Euro OverNight Index Average (EONIA), three and six-month Euro Interbank Offered Rate (EURIBOR) and the three and seven-year swap rate. Finally using these rates for the timeframe provides a total sample of 75 observations. Table 1.2 provides an overview for each retail interest rate with their corresponding market interest rate.

Table 1.2 Retail interest rates Corresponding Market interest rates

Overnight Deposits EONIA

Savings Deposits Three-month EURIBOR

Short-term time deposits Three-month EURIBOR

Long-term time deposits Three-year swap rate

Overdrafts EONIA

Short-term loans for consumer credit Six-month EURIBOR Long-term loans for consumer credit Seven-year swap rate Short-term loans for house purchase Three-month EURIBOR Long-term loans for house purchase Seven-year swap rate Short-term loans to non-financial corporations Three-month EURIBOR Long-term loans to non-financial corporations Seven-year swap rate

Source: ECB Monthly Bulletin August 2009

Error-Correction Model specification

To empirically determine the effectiveness of the interest rate pass-through effect in the Euro area a single equation ECM will be used. The use of a single equation ECM enables the determination of the speed and size of the adjustments of retail interest rates to market interest rates. The specific ECM that is used in the empirical study is defined as follows:

∆𝐵𝑅_! = 𝜑 + 𝛿 𝐵𝑅_!!!− 𝛽𝑀𝑅_!!! + 𝛼_!∆𝑀𝑅_!+ 𝛼_!∆𝑀𝑅_!!!+ 𝜂∆𝐵𝑅_!!!+ 𝜀_! In this single equation ECM the changes in the retail interest rates (∆𝐵𝑅!) are

explained by adjustments towards the long-run equilibrium between retail interest rates and market interest rates, as measured by the parameter 𝛿, which reflects the speed of adjustment. Furthermore ∆𝐵𝑅_! is determined by changes and lags in the relevant market rate (∆𝑀𝑅!, ∆𝑀𝑅!!!) and lagged changes in the retail interest rate

itself (∆𝐵𝑅_!!!). Finally, the error term 𝜀_! measures any omitted factors as well as all other factors that determine the level of the retail interest rate. Such as bank’s cost of intermediation, which contains interest rate risk, credit risk, risk aversion, operating costs, liquidity and product diversification. Additionally, parameters 𝛼_! measures the immediate pass-through and 𝛽 the final pass-through from market interest rates to retail interest rates.

(13)

In order to regress the single equation ECM mentioned earlier, the model must be slightly rewritten to the following model.

∆𝐵𝑅_! = 𝜑 + 𝛿𝐵𝑅_!!!+ 𝛾𝑀𝑅_!!!+ 𝛼_!∆𝑀𝑅_!+ 𝛼_!∆𝑀𝑅_!!!+ 𝜂∆𝐵𝑅_!!!+ 𝜀_! After regressing this specific model, the model can be converted back into the initial single equation ECM. However, this specific specification of the interest rate pass-through effect hinges on several conditions, one of which is that the retail and market interest rates don’t return to their past values, also referred to as being non-stationary. Additionally it must be established, whether these rates are cointegrated. E.g. a long-run stable relationship can be determined for these interest rates. Earlier studies have confirmed by using cointegration and standard unit root tests that the conditions for using the single equation ECM have been satisfied for retail and market interest rates. However for the short-term loans for consumer credit a long-run stable relationship with the six-month EURIBOR couldn’t easily be detected7. Furthermore, to avoid the assumption that the observations of the error term 𝜀_! have the same variance across the observation period a robust standard error regression is used. This will circumvent any heteroscedasticity issues inherit in the data. Despite the fact that the assumptions mentioned above have been proven to be satisfied, additional tests are performed to ensure that the variables used in the regression are non-stationary and cointegrated.

Dicky-Fuller unit root test

In order to conduct a regression using the single equation ECM mentioned earlier, the assumption of non-stationary variables must be satisfied. One of the most popular tests developed to determine whether a variable is non-stationary is the Dicky-Fuller (DF) unit root test. There are other tests to conduct a unit root test; however there is no uniformity for the best test and most of the times the results coincide, therefore the DF unit root test will be used. The general DF test involves fitting the following model using ordinary least squares (OLS):

𝑥! = 𝛼 + 𝜃𝑥!!!+ 𝛿𝑡 + 𝜀!

Where 𝜀_! is an independently and identically distributed (iid) zero-mean error term and 𝛼 and 𝛿 are initially set to zero.

(14)

This general DF test uses the 𝑍 𝑡 statistic, 𝑍 𝑡 =_!"(!)!!! , in testing the hypotheses. The hypotheses used in the DF test are 𝐻_!: 𝜃 = 1 against 𝐻_!: 𝜃 < 1 where the null corresponds to the variable having a unit root, or being non-stationary. However, fitting this general model will likely result in serial correlation. Fortunately, this issue can be circumvented using an augmented Dicky-Fuller (ADF) test. The ADF test extends the general model mentioned earlier by adding lagged terms of the implied variable, which results in the following model:

𝑥_!= 𝛼 + 𝜃𝑥_!!!+ 𝛾_!𝑥_!!!

!

!!!

+ 𝛿𝑡 + 𝜀_!

This fitted regression is used in determining whether the variables used in the ECM have unit root. Furthermore the ADF test is performed assuming a random walk with drift, but without a trend. Excluding a trend implies restricting 𝛿 = 0 in the

regression. Additionally, using an ADF test implies choosing a number of lagged terms8 (𝑘). The common rule of thumb for determining 𝑘 has been suggested by Schwert (1989) and has the following form: 𝑘!"# = 12 ∗ _!""!

! !

The Johansen test for Cointegration

The final assumption that needs to be satisfied to perform the single equation ECM regression is that the retail interest rates and their corresponding market interest rates are cointegrated. To determine whether these two variables are cointegrated, a so-called Johansen’s trace statistic method is used. This method is based on Johansen’s maximum likelihood (ML) estimator of the parameters of a cointegrating vector error correction model (VECM). The basic form of the VECM is:

Δ𝑥_! = 𝛼𝛽!_𝑦

!!!+ Γ!

!!!

Δ𝑦_!!!+ 𝜖_!

Where 𝑥 is a (K x 1) vector of variables integrated to the first level I(1), 𝜖! is a

(K x 1) vector of normally distributed errors that are serially uncorrelated but have a coexistent covariance matrix Ω, and 𝛼 and 𝛽 are (K x r) parameter matrices with rank r < K, Γ!… , Γ!!! are (K x K) matrices of parameters.

(15)

Based on the ML estimator the trace statistic can be derived, which has the following structure: −𝑇 ln (1 − 𝜆! ! !!!!! )

Where T is the number of observations and 𝜆_! are the estimated eigenvalues. The null hypothesis of the trace statistic is that there are no more than r cointegrating relations.

4. Results

For each variable used in the empirical study a DF unit root test is performed to determine whether or not the tested variable is non-stationary. The following table (1.3) is given to summarize the results for all variables.

Table 1.3 Z(t) p-value

Short-term loans for house purchase (STBR1) -0.565 0.2875 Short-term loans for consumer credit (STBR2) -3.184* 0.0013* Short-term loans to non-financial institutions (STBR3) -0.948 0.1739 Overdrafts non-financial institutions (STBR4) 0.343 0.6331

Overdrafts Households (STBR5) 0.321 0.6247

Overnight Deposits non-financial institutions (STBR6) -0.77 0.2225

Overnight Deposits Households (STBR7) 0.298 0.6166

Short-term time deposits non-financial institutions (STBR8) -1.089 0.1407 Short-term time deposits households (STBR9) -0.642 0.2618

Savings Deposits (STBR10) -0.268 0.3951

EONIA (STMR1) -1.218 0.1144

EURIBOR three-months (STMR2) -1.004 0.1601

EURIBOR six-months (STMR3) -0.809 0.2112

Long-term loans for house purchase (LTBR1) -0.443 0.3297

Long-term loans for consumer credit (LTBR2) 1.22 0.8859

Long-term loans to non-financial institutions (LTBR3) -1.105 0.1373 Long-term time deposits non-financial institutions (LTBR4) 0.05 0.5199

Long-term time deposits Households (LTBR5) -0.763 0.2246

Three-year swap rate (LTMR1) -2.481* 0.0083*

Seven-year swap rate (LTMR2) -2.214** 0.0157**

(16)

As seen from table 1.3 all variables except for STBR2, LTMR1 and LTMR2 are non-significant. This corresponds according to the DF unit root test to the following result. The variables, which are non-significant, are considered to be non-stationary or having unit root. Consequently, variables STBR2, LTMR1 and LTMR2 are stationary or not having unit root.

In addition to the DF unit root test, each bank interest rate and its corresponding market interest rate is tested for cointegration using the Johansen test. The following table (1.4) is given to summarize the results for all variables.

Table 1.4 Market Rate Trace statistic Maximum rank

STBR1 STMR2 11.2085 0

STBR2 STMR3 Not Applicable Full Rank

STBR3 STMR2 3.4454 1 STBR4 STMR1 1.1251 1 STBR5 STMR1 1.2297 1 STBR6 STMR1 0.0244 1 STBR7 STMR1 1.241 1 STBR8 STMR2 3.2017 1 STBR9 STMR2 2.2097 1 STBR10 STMR2 0.0092 1 LTBR1 LTMR2 13.8861 0 LTBR2 LTMR2 11.9296 0

LTBR3 LTMR2 Not Applicable Full Rank

LTBR4 LTMR1 Not Applicable Full Rank

LTBR5 LTMR1 3.1187 1

In the Johansen test for cointegration the maximum rank (r) corresponds to the amount of cointegrating equations. Johansen’s procedure for estimating 𝑟 is to accept 𝑟 for the first r for which the null hypothesis is not rejected. Therefore for all variables except for STBR1, LTBR1 and LTBR2 it can be said that they exhibit no

(17)

Error-Correction Model

After conduction both the DF unit root test and the Johansen test for cointegration the single equation ECM can be estimated using linear regression. As mentioned earlier a different linear model is regressed after which it is rewritten to reflect the correct ECM as proposed earlier. The following table (1.5 & 1.6) are given to summarize the results for all variables. Additional output for each variable can be found in the appendix for further reference.

Table 1.5 Immediate pass-through 𝜶𝟏 Final pass-through 𝜷 Speed of adjustment 𝜹 𝑹𝟐 Adj. 𝑹𝟐

STBR1 0.373* 1.012* -0.087* 0.463 0.424 STBR2 -0.999*** 0.112 -0.118* 0.187 0.128 STBR3 0.421* 0.966 0.015 0.846 0.835 STBR4 0.256* 4.840* -0.021 0.486 0.433 STBR5 0.120 4.186* -0.028 0.262 0.187 STBR6 0.173* 0.346* -0.192* 0.839 0.828 STBR7 0.0511* 0.267** -0.071** 0.789 0.774 STBR8 0.629* 0.679 -0.061 0.714 0.693 STBR9 0.200 3.612*** -0.014 0.537 0.504 STBR10 -0.118 0.734* -0.076* 0.575 0.544

*Significant at 1%; **Significant at 5%; ***Significant at 10%;

Table 1.6 Immediate pass-through 𝜶𝟏 Final pass-through 𝜷 Speed of adjustment 𝜹 𝑹𝟐 Adj. 𝑹𝟐

LTBR1 -0.003 0.074 -0.043* 0.365 0.319

LTBR2 -0.121 0.244 -0.065 0.168 0.107

LTBR3 0.275 0.536 -0.043 0.0283 -0.042

LTBR4 -0.254*** 0.451 -0.042 0.164 0.104

LTBR5 -0.187 0.352 -0.023 0.246 0.191

*Significant at 1%; **Significant at 5%; ***Significant at 10%;

5. Discussion

Campbell and Clarida (1987) as well as Newbol et al. (2001) have confirmed that nominal retail and market interest rates are often non-stationary. However, in some instances this might not be true due to several issues. In this particular empirical study it can be concluded that the variables STBR2, LTMR1 and LTMR2 are stationary, which implies they display a mean reverting process.

(18)

However, as several studies9 have pointed out, there are some issues related to the Dicky-Fuller unit root test that makes any inference based on these results less correct. In general, interest rates are highly serial correlated and in most of the cases the series used in empirical studies are near or fully integrated. Therefore rejecting the null hypothesis in the unit-root test might jus bet a result of deficiency in statistical power rather than evidence of stationary variables (Wu & Zang, 1996, p. 605). Additionally, the variables in question are short-term loans for consumer credit and both swap rates, which have been found to be more reluctant to structural changes in recent studies conducted by the ECB10. Therefore any stationarity in these variables could be a confirmation of these studies. Finally, it might also be the case that the harmonised data used to conduct the empirical study is inherently biased due to the financial turmoil in the period 2009 – 2015.

The Johansen test for cointegration is one of the most used tests in determining the cointegration of variables. However, this test relies on one very strict assumption. Namely, it relies on the assumption that the variables in question are unit root or non-stationary. Nevertheless, this assumption is often not satisfied, as has been shown in the previous DF unit root test. In this particular empirical study it can be concluded that the variables STBR1, LTBR1 and LTBR2 don’t display cointegrating equations, despite the fact that these variables are found to be non-stationary. However,

Hjalmarsson and Österholm (2007) point out that there seems to be a high probability of a doubtful rejection in the Johansen test due to near rather than pure unit root processes. Therefore, it might be the case that these variables display no cointegration due to this issue inherently present in the data. Furthermore, it might also be the case that the corresponding market rates as suggested by the ECB aren’t representative for these variables.

Error-Correction Model: short-term interest rates

As seen from table 1.5 almost all short-term interest rates respond with a slight lag in response to changes in their corresponding market interest rate.

9_{See Cochrane (1991), Campbell & Perron (1991) and Dejong et al. (1992)} 10_{See ECB (2002-2014)}

(19)

However, the short-term interest rate on loans to non-financial institutions (STBR3) seems to respond without this lag. This might be attributable to the fact that non-financial institutions have more bargaining power in choosing between different sources of funding and due to the significant size of funding in terms of money amount. Overall these findings are in correspondence with the findings from previous studies with regards to the sluggish speed of adjustment in short-term interest rates. Although these findings correspond to earlier studies, the significance and adjusted 𝑅!_{levels differ quite a lot among these interest rates. The size of the immediate}

pass-through effect differs significantly between these short-term interest rates. Overall these effects are positive, however for the interest rates on consumer credit and

savings deposits they seem to be negative. This might be explained by the fact that the short-term rate on consumer credit is a stationary variable and that immediate pass-through is not significant for the savings deposits. Overall it can be concluded that the findings with regards to the immediate through exhibit a less can complete pass-through in the short-term, which corresponds to earlier studies held after the financial crisis. Finally, the final pass-through seems to differ widely between the short-term rates. Only STBR1, STBR3, STBR8 and STBR10 seem to display a complete to near complete pass-through. All other variables seem to display large deviations. As mentioned earlier, this far from perfect final pass-through might be an indication of the distorting effects from the aftermath of the global financial crisis. As the negative effects are still ongoing the pass-through might be distorted for quite some while in the Euro area.

Error-Correction Model: long-term interest rates

As seen from table 1.6 the long-term interest rates also display a slight lag in response to changes in their corresponding market interest rates. However, in contrast to the short-term rates these long-term rates seem to respond slightly quicker. The immediate pass-through for these long-term interest rates ranges from low to

moderate, which coincides with the fact that in general the immediate effects on long-term rates are less significant compared to the short-long-term rates. Finally, the final pass-through from changes in market interest rates to long-term rates is far from perfect and even less complete when compared to the short-term rates.

(20)

Again this might be an indication of the distorting effects of the aftermath of the global financial crisis. However, it must be noted that the adjusted 𝑅!_{levels are}

extremely low in all cases.

6. Conclusion

The interest rate pass-through effect remains on of the most important channels through, which monetary policy tries to influence and stimulate the real economy. Earlier studies have shown that in the years previous to the global financial crisis, this so-called pass-through effect worked properly. However, more presently conducted studies have shown that this might not be the case in the Euro area for the years after the crisis. Empirical results from this study largely confirm these findings. Despite several differences among short-term and long-term interest rates in terms of their speed of adjustment to market interest rates, both short-term and long-term rates seem to display a universal sluggish response to changes in their corresponding market interest rates. Additionally, the immediate pass-through effect for both the short and long-term rates displays either a positive of negative lag. More interestingly, in almost all cases the final pass-through seems heavily distorted, either by an extremely high or low effect. Taking these finding into account, it can be concluded that traditional pass-through effect is distorted and therefore not properly working in the Euro area. However, it must be noted that further research is required to fully determine whether this is the case for all interest rates or whether these findings fully reflect the whole spectrum.

(21)

7. References

Angeloni, I., Kashyap, A. K., Mojon, B., & Terlizzese, D. (2003). Monetary Transmission in the Euro area: does the interest rate channel explain all? National

Bureau of Economic Research , 1-38.

Bernanke, B. S., & Blinder, A. S. (1992). The Federal Funds Rate and the Channels of Monetary Transmission. The American Economic Review , 901-921.

Bernanke, B. S., & Gertler, M. (1995). Inside the Black Box: the credit channel of monetary policy transmission. National Bureau of Economic Research , 1-47. Campbell, J. Y., & Clarida, R. H. (1987). The term structure of euromarket interest rates: An empirical investigation. Journal of Monetary Economics , 25-44.

Campbell, J. Y., & Perron, P. (1991). Pitfalls and opportunities: What

macroeconomists should know about unit roots. National Bureau of Economic

Research , 141-220.

Cochrane , J. H. (1991). A critique of the application of unit root tests. Journal of

Economic Dynamics and Control , 275-284.

Cottarelli , C., & Kourelis, A. (1994). Financial Structure, Bank Lending Rates, and the Transmission Mechanism of Monetary Policy . International Monetary Fund , 587-623.

Darracq Paries, M., Moccero, D. N., Krylova, E., & Marchini, C. (2014, August). The retail bank interest rate pass-through: The case of the Euro area during the financial and sovereign debt crisis. ECB Occasional Paper Series.

De Bondt, G. (2002). Retail bank interst rate pass-through: new evidence at the Euro area level. European Central Bank , 1-43.

DeJong, D. N., Nankervis, J. C., Savin, N. E., & Whiteman, C. H. (1992). The power problem of unit root tests in time series with autoregressive errors. Journal of

(22)

Engle, R. F., & Granger, C. W. (1987). Co-Integration and Error Correction: Representation, Estimation, and Testing. Econometrica , 251-276.

Gigineishvili, N. (2011). Determinants of interest rate pass-through: Do

Macroeconomic conditions and financial markets matter? International Monetary

Funds , 1-18.

Hjalmarsson, E., & Österholm, P. (2007). Testing for Cointegration Using the Johansen Methodology when Variables are Near-Integrated. International Monetary

Fund.

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 , 104-119.

Johansen, S. (1991). Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models. Econometrica , 1551-1580.

Johansen, S., & Juselius, K. (1990). Maximum likelihood estimation and inference on cointegration- with applications to the demand for money. Oxford Bulletin of

Economics and Statistics , 170-210.

Kashyap, A. K., & Stein, J. C. (2000). What Do a Million Observations on Banks Say About the Transmission of Monetary Policy? The American Economic Review , 407-427.

Leybourne, S., Newbold, P., & Vougas, D. (1998). Unit roots and smooth transitions.

Journal of time series analysis , 83-97.

Liu, M.-H., Margaritis, D., & Tourani-Rad, A. (2008). Monetary policy transparency and pass-through of retail interest rates. Journal of Banking & Finance , 501-511. Lombardi, M., & Illes, A. (2013, September). Interest rate pass-through since the financial crisis. BIS Quarterly Review , pp. 57-66.

Mishkin, F. S. (1996). The channels of Monetary transmission: lessons for monetary policy. National Bureau of Economic Research , 1-28.

(23)

Sørensen, C. K., & Werner, T. (2006). Bank interest rate pass-through in the Euro area: a cross country comparison. The European Central Bank , 1-65.

Saborowski, C., & Weber, S. (2013). Assessing the Determinants of Interest Rate Transmission Through Conditional Impulse Response Functions. International

Monetary Fund , 1-16.

Sander, H., & Kleimeier, S. (2004). Interest rate pass-through in an enlarged Europe: The role of banking market structure for monetary policy transmission in transition countries. Journal of Economic Literature , 1-66.

Schlüter, T., Hartmann-Wendels, T., Sievers, S., & Busch, R. (2012). Determinants of the interest rate pass-through of banks − evidence from German loan products.

Deutsche Bundes Bank , 1-72.

The Bank of England. (2005). The transmission mechanism of monetary policy. The

Bank of England , 1-12.

The European Central Bank. (2000, July). Monetary policy transmission in the euro area. ECB Monthly Bulletin , pp. 43-58.

The European Central Bank. (2009, August). Recent developements in the retail bank interest rate pass-through in the Euro area. ECB Monthly bulletin , pp. 93-105.

Tong, H., & Wei, S.-J. (2008). Real Effects of the Subprime Mortgage Crisis: Is it a Demand or a Finance Shock? International Monetary Fund , 1-37.

Weth, M. A. (2002). Determinants of the interest rate pass-through of banks − evidence from German loan products. Deutsche Bundes Bank , 1-40.

Wu, Y., & Zhang, H. (1996). Mean reversion in interest rates: New evidence from OECD Countries. Journal of Money, Credit and Banking , 604-612.

(24)

8. Appendix

Output: ECM regression estimates for short-term rates

Table 1.7 𝜹 𝛾 𝜶𝟏 𝜶𝟐 η 𝑹𝟐 Adj. 𝑹𝟐 STBR1 -0.087* (0.028) 0.089* (0.025) 0.373* (0.118) -0.030 (0.103) 0.036 (0.113) 0.463 0.424 STBR2 -0.118* (0.044) -0.013 (0.057) -0.998*** (0.572) 0.454 (0.477) -0.175 (0.108) 0.187 0.128 STBR3 0.015 (0.028) 0.014 (0.014) 0.421* (0.087) 0.148 (0.098) 0.249** (0.099) 0.846 0.835 STBR4 -0.021 (0.027) 0.103* (0.026) 0.256* (0.068) -0.012 (0.073) 0.039 (0.139) 0.486 0.433 STBR5 -0.028 (0.027) 0.119* (0.033) 0.120 (0.085) -0.016 (0.088) -0.175 (0.143) 0.262 0.187 STBR6 -0.192* (0.047) 0.066* (0.019) 0.173* (0.032) -0.041 (0.037) 0.248* (0.100) 0.839 0.828 STBR7 -0.071** (0.029) 0.019** (0.009) 0.051* (0.019) 0.026 (0.019) 0.425* (0.089) 0.789 0.774 STBR8 -0.061 (0.040) 0.042 (0.031) 0.629* (0.144) 0.187 (0.138) -0.037 (0.118) 0.714 0.693 STBR9 -0.014 (0.025) 0.049*** (0.027) 0.200 (0.190) 0.395** (0.173) 0.170 (0.111) 0.537 0.504 STBR10 -0.076* (0.028) 0.056* (0.017) -0.118 (0.087) 0.297* (0.075) 0.110 (0.094) 0.575 0.544 *Sig. 1% **Sig. 5% ***Sig. 10%

Output: ECM regression estimates for long-term rates

Table 1.8 𝜹 𝛾 𝜶𝟏 𝜶𝟐 η 𝑹𝟐 Adj. 𝑹𝟐 LTBR1 -0.043* (0.011) 0.003 (0.007) -0.003 (0.0192) -0.013 (0.016) 0.191 (0.115) 0.365 0.319 LTBR2 -0.065 (0.053) 0.016 (0.023) -0.121 (0.075) -0.016 (0.065) -0.281** (0.112) 0.168 0.107 LTBR3 -0.043 (0.070) 0.023 (0.067) 0.275 (0.215) -0.003 (0.187) 0.075 (0.139) 0.0283 -0.042 LTBR4 -0.042 (0.028) -0.019 (0.037) -0.254*** (0.129) 0.240** (0.115) 0.084 (0.111) 0.164 0.104 LTBR5 -0.023 (0.021) -0.008 (0.027) -0.187** (0.098) 0.192** (0.086) 0.320* (0.107) 0.246 0.191 *Sig. 1% **Sig. 5% ***Sig. 10%

An empirical study on the interest rate pass-through effect in the euro area

Faculty of Economics & Business