1
Combined Master’s Thesis
MSc Finance & MSc EconomicsEffect of Monetary Policy on Bank
Risk-Taking in the Eurozone
I.F.A. (Irma) Ridder1
Faculty of Economics and Business, University of Groningen
Supervised by: dr. C.G.F. (Christiaan) van der Kwaak
January 19th, 2018
Abstract: This paper examines the effect of the monetary policy stance of the European Central Bank (ECB) on
bank risk-taking within the Eurozone. The post-crisis low interest rate set by the ECB might enhance this relationship. The paper uses an unbalanced dataset which includes 106 banks in the Eurozone within the time period of 2002-2016. Results show a negative relationship between the Taylor rule residual and the ratio of risk-weighted assets to total assets in certain specifications of the model. This indicates that the policy of the ECB, which decreases the interest rate, leads to an increase in risk-taking of Eurozone banks. Results are also found with respect to the Asset Purchasing Programme (APP) of the ECB. The APP relates negatively to risk-taking by banks, indicating that the APP decreases risk-taking by banks in the Eurozone.
Course code: EBM000A20
Keywords: Bank risk-taking, interest rates, European Central Bank JEL codes: E43, E58, G21, G28
1 Student number: S2350645
2
1. Introduction
The current monetary policy of the European Central Bank (ECB) has led to low interest rates. These low interest rates have dropped to previously unthinkable negative values for deposit rates. The utmost "zero lower bound" has first been broken by the ECB in June 2014. By doing so, the ECB responded to the aftermath of The Great Recession and European Sovereign Debt Crisis in which inflation was low and growth became sclerotic. The policies conducted by the central bank are supposed to lead inflation rates to approach the bank its objective. Therefore, the policies set by the ECB were needed to provide positive monetary stimulus to the Eurozone. If not pursued, the inflation rate was assumed to further diminish, causing activity and employment to stall. To complement the low interest rates and thus ensure that the targets of the ECB are met, the bank launched the Asset Purchasing Programme (APP). With the APP, the ECB is involved in purchasing private sector securities and public sector securities. The APP leads to an increase in security prices and an increase in the money base. These two effects result into a fall of interest rates and therefore a decrease in loan prices. This stimulates the loan demand, hence the economy. Moreover, as prices rise, the inflation rate will approach to its target (Cœuré, 2016).
3 Previous research finds evidence which supports the arguments made by the critics. Most papers find a negative relation between interest rates and the risk-taking behaviour of banks (Delis and Kouretas, 2011, Apel and Claussen, 2012, Cociuba et al., 2016). These results indicate that the monetary policy of the ECB could cause a deterioration of financial stability. A decrease in financial stability can have a great effect on the economy as a whole (Rajan, 2006). These consequences might entail that the current interest rate could be the foundation for a new crisis if banking supervision is insufficient. The rates have been below the physical zero lower bound already, increasing the urgency to review whether changes in the interest rate will cause consequences for bank risk-taking and thereby financial stability. Specifically, the paper answers the question: What is the relationship between the monetary policy stance of the European Central Bank and bank risk-taking within the Eurozone?
This paper combines the best-practises of different papers to investigate the relation between the low interest rate and bank risk-taking behaviour (Delis & Kouretas 2011; Dell’Arricia et al., 2017). The model estimates the relationship between the Taylor rule residual as a measure of interest rates and risk-weighted assets to total assets as a measure of bank risk-taking. The paper measures the effect of interest rate via the Taylor rule residuals to address endogeneity problems. The Taylor rule residual is estimated as the difference between the EONIA rate and the Taylor rule (Taylor, 1993). Next to the main relationship, the paper estimates the effect of quantitative easing on bank taking. Quantitative easing is expected to influence bank risk-taking because if the interest rates decrease, banks are more likely to seek for alternative investments, hence taking more risks. To the best of my knowledge, there has not been a paper which reviews both the effect of the interest rate and the asset purchasing programme of the ECB. This paper contributes to the literature by adding a measure of APP to the model of Delis and Kouretas (2011). Also, this paper measures the effects over a longer time series compared to Delis and Kouretas (2011).
4 The remainder of the paper is organized as follows. Section 2 discusses previous research that has been conducted on this topic. Section 3 elaborates on the model used to obtain the results. Section 4 provides data specifications and Section 5 discusses the empirical results. Section 6 concludes the paper, provides policy implications and offers suggestions for further research.
2. Literature review
Before diving into the theory written on risk-taking behaviour, it is important to understand risk-shifting behaviour of financial institutions. As argued by Jensen and Meckling (1976) and Myer (1977), risk-shifting leads to excessive risk-taking due to limited liabilities of debt issuers. Limited liabilities arise because management of financial institutions are not liable for the losses encountered when investments fail. The limited liability causes debt issuers ignoring the downside risk of an investment and them only focusing on the upside potential (Dubecq et al., 2010).
As for the effect of interest rates on risk-shifting by financial institutions, Rajan (2006) hypothesizes that low interest rate environments amplify the risk-shifting in financial markets. Risk-shifting leads financial institutions to increase their risk-taking because in low interest rate environments the safe bonds do not provide enough profits for survival. Therefore, risk-shifting causes them to forget the downside risk of riskier assets because they are solely focused on the upside potential. A second form of risk-shifting relates mostly to hedge funds. Hedge funds receive little return on their assets when they take little risks and therefore boost risk-taking behaviour. Therefore, hedge funds borrow more in order to take more risks in an attempt to increase their profits. This behaviour increases the amount of risky assets as well as the leverage of hedge funds.
5 conclude that until that moment, central banks have not considered the impact of monetary policy on the pricing of risks.
Apel and Claussen (2012) describe another channel through which monetary policy affects the risk appetite of banks. This channel is named the collateral channel. If the interest rate decreases, the value of assets increases because the value of future cash flows increases. Hence, banks are more willing to lend because the value of the collateral increases. This could lead to banks behaving differently than in an environment where interest rates are higher, while banks are more willing to take on risks. Therefore, the risk-taking behaviour is amplified by interest rate fluctuations. However, Apel and Claussen (2012) are critical about the amount of influence that monetary policy actually has on bank taking. They state that banks adjust their taking based on the long-term perspectives. Therefore, the interest rates influencing bank risk-taking are long-term interest rates. They state that central banks can only directly affect short-term interest rates, meaning that the relationship between the interest rate and bank risk-taking might be caused by other factors than monetary policy. Therefore, investigating solely the effect that the interest rate has on bank risk-taking might not cover the effect of monetary policy on bank risk-taking. The paper thus recommends studying the effect between expansionary monetary policy and risk-taking.
6 However, the theoretical relationship between the interest rate and the bank risk-taking behaviour is not as straightforward as the papers described above suggested. Dell’Aricca et al. (2013) notice that the effect of real interest rates depends on the capital structures of banks. The paper argues that if banks are able to change their capital structure, a fall in interest rates guides banks to become more leveraged and more risky. However, when capital structures are fixed, low leveraged banks increase their risks. The contrary holds for well-leveraged banks.
Several papers have found evidence which empirically supports the channels described above. The paper by Delis and Kouretas (2011) describes the relationship between various measures of interest rates and bank risk-taking. For bank risk-taking, they use the measures of the ratio of risk-weighted assets to total assets and the ratio of non-performing loans to total loans. The paper focuses on Eurozone data and finds that the low interest rates lead to banks having lower margins and fewer information asymmetries. Information asymmetries are present because the bank does not fully know the creditworthiness of its clients. When interest rates are low, clients have less incentives to lie about their creditworthiness. These effects drive banks to lower their lending standards and hence increase their risky assets. The paper finds robust negative relationships between interest rates and bank risk-taking.
The findings of Özşuca and Akbostanci (2016) also provide evidence in favor of the negative relationship between interest rates and bank risk-taking. They use a database focused on Turkey and use different measures of risk-taking behaviour: non-performing loans ratio, z-index and standard deviation of the return on assets. The paper finds evidence of a risk-taking channel in their sample. Concretely, they conclude that low short-term interest rates decline risks on already existing loans. However, banks do increase risk-taking in low short-term interest rate environments.
7 Thus, the important finding is that when supervision is weak, it enhances the main relationship. Note that banking supervision is the responsibility of national governments in the Eurozone.
As Heins (2017) describes in his paper, most papers written in the past focus on measurements of bank risk-taking which evaluate the risk exposure of banks instead of their willingness to take on risks. To overcome this, the author proposes to use two proxies of risk-taking behaviour. Namely, the use of loan growth and the ratio risk-weighted assets to total assets. Estimating the interest rate by using the Taylor rule residual, the paper finds evidence in favour of the risk-taking channel. The Taylor rule residual is also used in estimating the interest rate in the paper of Delis et al. (2017). The paper finds a negative weak relationship between bank risk-taking, measured as corporate loan spreads, and interest rates in the pre-crisis period. The paper estimate Taylor rule residuals by regressing the shadow federal funds rate on the output gap and the inflation gap. The shadow federal funds rate is the policy interest rate in conventional monetary policy environments. The paper urges future research to focus on the coordination of monetary and regulatory policies.
Another approach to measuring the effect of bank risk-taking has been applied by Paligorova and Santos (2017). In their paper, they use a dataset composed of Senior Loan Officers Opinion Survey from the United States. The paper hypothesizes that in times of monetary easing, assuming no change in loan quality, banks regard loans to be less risky and thus supply more risky loans. Thereby, banks that are more risk-loving will provide more risky loans when experiencing loose monetary policy. Results show that in times of monetary policy easing, loan spreads for riskier firms decrease which leads to an underpricing of corporate loan risk and an increase in the supply of riskier loans.
8 emerging markets. The quality instead of quantity measure is also used by Dell’Ariccia et al. (2017). They suggest that the focus of research should be more on the quality of bank credit. They measure bank risk-taking as the risk rating of new loans and find a negative relationship between this measure and short-term interest rates. The relationship is stronger in times of financial distress and for banks that have lower capital holdings. As a robustness test, the paper uses the Taylor rule residual to counter any possible endogeneity problems.
The paper by Oliveira de Moraes et al. (2016) attempts to bridge the microprudential view with the macroprudential view. The paper notices that banks change their capital holdings in response to changes in monetary policy. More specifically, changes in monetary policy alter the capital adequacy ratio and the amount of provisions that banks hold. When interest rates and reserve requirements decrease, banks lower their provisions and become thus less solvent, which decreases the capital adequacy ratio. These findings suggest that banking supervisors should take into account the monetary policy set by central bankers.
Similar results are also found by Adesina and Mwamba (2016) who estimate the relationship between common equity capital and risk-taking behaviour for banks in South Africa. Three measures of risk-taking are employed: the ratio of loan loss reserves to total assets, Z-score and the ratio of non-performing loans to total loans. They find robust results which indicate that the higher the common equity capital within a bank, the lower the risk-taking behaviour. This indicates that a sufficient level of common equity capital mitigates the risk-taking behaviour of banks. Thereby, they state that Basel III requirements should be able to lead to a more stable banking sector in South Africa.
9
3. Methodology and hypotheses
I will use an adjusted model based on the paper by Delis and Kouretas (2011). They regress a risk variable upon a constant, interest rate, macroeconomic control variables, bank-specific control variables and an error term. The research will be based on an unbalanced dataset with annual data from 2002-2016. Data is obtained from 2002 onward while in that year the Euro cash changeover happened. Therefore minimizing differences between the countries in the analysis. The dataset is corrected for banks in countries which joined the Eurozone at a later date. Data is included for banks from Slovenia (2007), Slovakia (2009) and Estonia (2011) from the year they joined onwards. The years in which country joined the Eurozone is specified in Appendix A. Also, due to data limitations as further explained in Section 4, Eurozone countries Cyprus, Malta, Latvia and Lithuania are deleted from the dataset. The model to estimate the hypotheses is:
𝑅𝑊𝐴𝑇𝐴𝑖,𝑡 = 𝛼𝑖+ 𝛽1𝑇𝑅𝑅𝑐,𝑡+ 𝛽2𝑄𝐸𝑡 + 𝛽3𝑀𝐶𝑐,𝑡+ 𝛽4𝑇𝑅𝑅𝑐,𝑡∗ 𝐶𝐴𝑅𝑖,𝑡 + 𝛽5𝐵𝐶𝑖,𝑡 + 𝜀𝑖,𝑡
The subscript 𝑖 is an indicator for bank level, 𝑡 is an indicator for time and 𝑐 is an indicator for country level. The variables are specified as follows. 𝑅𝑊𝐴𝑇𝐴𝑖,𝑡 is the risk-taking variable, for which the ratio risk-weighted assets to total assets will be used. 𝛼𝑖 represents the constant. 𝑇𝑅𝑅𝑐,𝑡 is the interest rate measure, for which the Taylor rule residual will be used. 𝑄𝐸𝑡 is the quantitative easing variable, for which the ratio quantitative easing to aggregate Eurozone GDP will be used. 𝑀𝐶𝑐,𝑡 indicates the macroeconomic control variables. 𝑇𝑅𝑅𝑐,𝑡∗ 𝐶𝐴𝑅𝑖,𝑡 is the interaction term between the Taylor rule residual and the capital adequacy ratio – tier 1. 𝐵𝐶𝑖,𝑡 are the bank control variables. 𝜀𝑖,𝑡 represents the error term.
10 the extended period of below target inflation. The APP is supposed to help guard against mispricing of risks within the financial sector (European Central Bank, 2017). The APP is part of the targeted longer-term refinancing operations of the ECB thereby reducing long–term interest rates (Diez de los Rios & Shamloo, 2017). Expectations are thus that the unconventional policy of the ECB positively affect the dependent variable. The interaction term is included because I believe that the capital ratio could mitigate the effect of interest rates on bank risk-taking based on the risk-shifting theory described above. An increase in capital ratio leads to more reserves and therefore pushes banks to less risk-taking behaviour. When interest rates are low, financial institutions increase their risk-taking (Rajan, 2006) as described above. By including the interaction variable, I should be able to measure the effect of bank regulations on the relationship between monetary policy and bank regulations. Expectations are that the interacting term is negative. When the capital ratio increases, the negative effect of interest rate on bank risk-taking should decrease.
11 Previous research has shown that there are two potential endogeneity problems with the equation described above (Delis and Kouretas, 2011). The problems exist with regard to various bank-specific control variables and the interest rate. The endogeneity problems with respect to the bank-specific control variables are independent of the construction of the variables and therefore handled in the robustness checks in Section 5. To give a preview, the potential issue is solved by using an instrumental variable regression with lagged variables for the endogenous bank-control variables. The second endogeneity problem exists while there might exist reverse causality between the interest rate and bank risk-taking. Above is described the effect that the interest rate has on bank risk-taking. However, bank risk-taking also affects the interest rate. The ECB its main goal is to maintain stable prices in order to increase economic welfare and the growth potential of an economy. Bank risk-taking leads to a deterioration of economic welfare and therefore influences the interest rates. To counter the problem of endogeneity with the interest rate variable, the Taylor rule residual is estimated using the method described next.
To measure short-term interest rates, it is appropriate to use either the nominal overnight rate (EONIA) or the Taylor rule residual (Maddaloni and Peydró, 2011). As stated before, the regression is likely to have a problem of endogeneity. To solve for this, the paper will use the Taylor rule. The Taylor rule is a model which determines the level at which central banks should set the nominal short-term interest rate as a result of to changes in economic conditions such as inflation and output (Taylor, 1993). In this paper, the Taylor rule will be calculated for each country (𝑐) and for each year (𝑡):
𝑖𝑐,𝑡∗ = 𝑟∗+ 𝜋𝑐,𝑡+ 𝛼𝜋(𝜋𝑐,𝑡− 𝜋∗) + 𝛼𝑦(𝑦𝑐,𝑡− 𝑦𝑐,𝑡∗ )
In this equation, 𝑖𝑐,𝑡∗ is the interest rate according to theory, 𝑟∗ is the equilibrium real interest rate (assumed to be 2%), 𝜋𝑐,𝑡 − 𝜋∗ shows the difference in inflation rate between the actual inflation rate and the natural inflation rate (𝜋∗) (as targeted by the ECB at 2%), 𝑦
𝑐,𝑡− 𝑦𝑐,𝑡∗ shows the percentage deviation of real GDP growth from the target (Dell’Ariccia, et al., 2017). The target GDP growth rate is calculated from potential GDP. The growth rate is for both real GDP and potential GDP calculated as the annual growth rate compared to the prior year. According to theory, 𝛼𝜋 and 𝛼𝑦 are equal to 0.5 (Taylor, 1993). Once calculated the 𝑖𝑐,𝑡∗ for each country in each year, the Taylor rule residual is obtained by calculating Equation (3). 𝑖𝑐,𝑡 is the EONIA rate. The Taylor rule residual will be measured in percentages.
12 𝑇𝑅𝑅𝑐,𝑡 = 𝑖𝑐,𝑡− 𝑖𝑐,𝑡∗
The model states that when the EONIA is below the as theory suggested interest rate, a negative Taylor rule residual will prevail. This would indicate that the EONIA is below the threshold as the interest rate should be set based on output and inflation rates. Based on Cociuba et al. (2016), the expectations are that when the short-term interest rate is the below theory predicted rate, the economy is in a very low short-term interest rate environment. Therefore, banks will increase their risk-weighted assets over total assets. The hypothesis is that the relationship between the Taylor rule residual and the ratio risk-weighted assets to total assets is likely to be negative.
Besides the effect of the interest rate, the equation will test what the effect of the APP is on bank risk-taking. The measure for this is the ratio quantitative easing to current level GDP (𝑄𝐸𝐺𝐷𝑃𝑡). The current level of GDP is the aggregate current GDP level of the Eurozone. The APP is supposed to have a positive effect on bank risk-taking. The variable is measured in percentages. The 𝑄𝐸𝐺𝐷𝑃𝑡 is only included from 2014 onwards, as the programme started in October 2014. For the rest of the years, 0 has been included as a data point for this variable. Expected is that the 𝑄𝐸𝐺𝐷𝑃𝑡 negatively affects bank risk-taking. As explained before, 𝑄𝐸𝐺𝐷𝑃𝑡 leads to higher loan demand and lower interest rates. Hence, to more risk-taking by banks.
13 (Jones, 2017). According to this idea, a negative effect of inflation is expected on bank risk-taking. However, inflation can increase the income margins of banks, therefore a positive effect of inflation on bank risk-taking might be expected (Bikker and Vervliet, 2017). 𝐷𝐶𝐺𝐷𝑃𝑐,𝑡 is expected to have a positive influence on bank risk-taking. as the amount of domestic credit provided by the banking sector increases, competition and developments are supposed to increases alongside it, leading to more risk-taking by banks (Delis and Kouretas, 2011)
To control for bank specifics, as best practises of Delis and Kouretas (2011) show, I use the following bank-specific control variables: capital adequacy ratio – tier 1 (𝐶𝐴𝑅𝑖,𝑡), total equity to total assets (𝑇𝐸𝑇𝐴𝑖,𝑡), lagged ratio profits before tax to total assets (𝑃𝐵𝑇𝑇𝐴𝑖,𝑡−1), natural logarithm of total assets (𝐿𝑂𝐺𝑇𝐴𝑖,𝑡) and ratio operating revenue to operating expenses (𝑂𝑅𝑂𝐸𝑖,𝑡).
𝐶𝐴𝑅𝑖,𝑡 is included to measure the effect of Basel capital requirements on bank risk-taking. I expect 𝐶𝐴𝑅𝑖,𝑡 to influence bank risk-taking while an increase in capital requirements should decrease bank risk-taking. 𝑇𝐸𝑇𝐴𝑖,𝑡 is included because the leverage of banks impacts bank risk-taking. As implied by the risk-sharing theory described before, low leveraged banks take less risks than high leveraged banks (Dell’Arricia et al., 2013). Therefore, the hypothesis is that 𝐶𝐴𝑅𝑖,𝑡 and 𝑇𝐸𝑇𝐴𝑖,𝑡 have a negative relationship to 𝑅𝑊𝐴𝑇𝐴𝑖,𝑡. The 𝐶𝐴𝑅𝑖,𝑡 and 𝑇𝐸𝑇𝐴𝑖,𝑡 are both included while 𝐶𝐴𝑅𝑖,𝑡 specifically shows the effect of regulatory requirements on bank risk-taking and 𝑇𝐸𝑇𝐴𝑖,𝑡 specifically shows the effect of leverage on bank risk-taking. 𝑃𝐵𝑇𝑇𝐴𝑖,𝑡−1 is included because profits in a period influences the risk-taking in the next period. If profits in time 𝑡 − 1 are high/low, this might lead to respectively more/less risk-taking in time 𝑡. Therefore, the hypothesis is that 𝑃𝐵𝑇𝑇𝐴𝑖,𝑡−1 has a positive impact on risk-taking. To correct for the size of banks, 𝐿𝑂𝐺𝑇𝐴𝑖,𝑡 is included. I expect size to have a negative effect on bank risk-taking while when the size of banks increase, banks become more risk averse. Lastly, to control for the efficiency of banks, 𝑂𝑅𝑂𝐸𝑖,𝑡 is included as a control variable. Banks which have higher profits are also more likely to have risky assets. Therefore, expected is that 𝑂𝑅𝑂𝐸𝑖,𝑡 relates positively to the ratio risk-weighted assets to total assets.
14 total revenue to total expenses. In this paper, I will use a ratio of operating revenue to operating expenses. The reason for this is that the variable total expenses was not available via Datastream. Lastly, Delis and Kouretas (2011) uses regulatory indexes as control variables. I do not included the indexes in the main analysis because the regulatory indexes only have data available for three years, meaning that a large majority of the dataset is omitted when performing the regressions. Next to this, there are collinearity problems between the indexes and quantitative easing to aggregate GDP which means that if the indexes are included, quantitative easing drops from the specification. I will elaborate on this issue in Section 5. As a control, the robustness check will include the indexes.
To sum up, the paper will investigate the following hypotheses:
Hypothesis 1: The Taylor rule residual negatively influences the ratio risk-weighted assets to total assets.
Hypothesis 2: Quantitative Easing positively influences the ratio risk-weighted assets to total assets.
4. Data and descriptive statistics
15 Table 1 gives a description of the data as specified in the regressions. The sample is not corrected for outliers to accurately reflect the full sample. This explains the values larger than 1 for the maximum on 𝑅𝑊𝐴𝑇𝐴𝑖,𝑡 where 12 of the 899 observations have a value greater than 1. This can be due to low credit ratings. If a claim has a credit rating below B-, then according to Basel The First Pillar – Minimum Capital Requirements, the banks have to hold 150% for claims on sovereigns, banks and corporates (Hull, 2015). On the contrary, the mean and median are very similar indicating few outliers. In the robustness checks, I will correct the sample for these outliers. When observing the 𝑇𝑅𝑅𝑐,𝑡, on average the EONIA (𝑖𝑐,𝑡) is 1.25% lower than the interest rate according to theory. On average, the 𝐶𝐴𝑅𝑖,𝑡 is 10.3%, which is well above the current minimum requirement for tier 1 capital as set by the Basel Committee of 6% according to Basel III (Basel Committee on Banking Supervision, 2011). Some values are negative for both 𝑇𝐸𝑇𝐴𝑖,𝑡, and 𝐶𝐴𝑅𝑖,𝑡, which could be due to banks having many bad loans on their balance sheet. Negative values for those measures would indicate insolvency by banks. If the bank removes those bad loans, 𝑇𝐸𝑇𝐴𝑖,𝑡 and 𝐶𝐴𝑅𝑖,𝑡can drop to negative values. For 𝑔𝑔𝑑𝑝𝑐,𝑡 the minimum value of -9.13% is obtained by Estonia in 2009 due to the financial crisis. The maximum value of 25.56% belongs to Ireland in 2015 which experienced a large increase due to a tax inversion. In 2015 a large bulk of companies changed to Ireland as their home country after merger and acquisitions increasing the GDP of Ireland tremendously (Inman, 2016).
Table 2 shows the correlation coefficients for all variables, except 𝑄𝐸𝐺𝐷𝑃𝑡due to the limited amount of data, used in the regression. Especially the correlation between the 𝑃𝐵𝑇𝑇𝐴𝑖,𝑡 and 𝑂𝑅𝑂𝐸𝑖,𝑡 is very high, 0.951, which is logical because the variable profits before tax exists of operating revenue and operating expenses. I will deal with this problem differently than Delis and Kouretas (2011) did, as explained in next section. Appendix C includes descriptive statistics on country level.
16
Table 1: Descriptive statistics
Standard
Variable Mean Median Min Max Deviation N
RWATA 0.553 0.570 0.000 2.909 0.246 899 TRR (%) -1.245 -1.308 -8.411 4.903 1.750 1543 CAR (%) 10.341 9.610 -7.300 57.330 4.341 955 TETA 0.089 0.065 -0.043 0.999 0.117 1361 QEGDP (%) 1.345 0.000 0.000 13.688 3.649 1543 ggdp (%) 0.899 1.092 -9.132 25.557 2.718 1543 infl (%) 1.765 1.900 -1.700 5.500 1.248 1528 PBTTA 0.006 0.006 -0.244 0.323 0.021 1361 LOGTA 17.101 17.074 10.644 21.509 2.099 1362 OROE 1.393 1.131 0.000 52.303 2.809 1246 DCGDP (%) 142.035 138.100 62 249 36.947 1543
Notes: The table reports the descriptive statistics for the variables used in the regression analysis. The variables
are respectively: RWATA is the ratio of risk-weighted assets to total assets, TRR is the Taylor Rule residual, CAR is the capital adequacy ratio, TETA is the ratio of total equity to total assets, QEGDP is the ratio of quantitative easing to aggregrate GDP for the Eurozone, ggdp is the growth rate of GDP, infl is the inflation rate, PBTTA is the ratio of profits before tax to total assets, LOGTA is the natural logarithm of total assets, OROE is the ratio of operating revenue to operating expenses, DCGDP is the domestic credit provided by the banking sector as a share of GDP.
17
Table 2: Correlation Coefficients
RWATA TRR CAR TETA ggdp infl PBTTA LOGTA OROE DCGDP
RWATA 1.000 TRR -0.183 1.000 CAR -0.230 0.414 1.000 TETA 0.438 0.056 0.239 1.000 ggdp 0.043 -0.616 -0.172 0.073 1.000 infl 0.191 0.582 0.186 0.258 0.029 1.000 PBTTA 0.185 -0.312 -0.045 0.335 0.509 0.143 1.000 LOGTA -0.385 -0.124 -0.096 -0.371 0.068 -0.205 -0.050 1.000 OROE 0.237 -0.306 -0.059 0.377 0.502 0.147 0.951 -0.107 1.000 DCGDP -0.123 0.406 0.079 -0.093 -0.411 0.150 -0.360 0.206 -0.368 1.000
Notes: The table reports the correlation coefficients for the variables used in the regression analysis. The variables are respectively: RWATA is the ratio of risk-weighted assets
18
5. Results
The following section will explain the results found by using the econometric model as described in Section 3. This section will describe the relations found in the model. Besides, it will verify the main results found by performing some robustness checks.
Before discussing the results, unit root tests show that 𝑇𝐸𝑇𝐴𝑖,𝑡, 𝐶𝐴𝑅𝑖,𝑡 and 𝐷𝐶𝐺𝐷𝑃𝑐,𝑡 are non-stationary. This is logical to understand for 𝑇𝐸𝑇𝐴𝑖,𝑡 and 𝐶𝐴𝑅𝑖,𝑡 while due to an increase in regulations, capital should increase over the years. For 𝐷𝐶𝐺𝐷𝑃𝑐,𝑡, the share of domestic credit provided by financial sector to GDP, has experienced an on average increase over the time period 2002-2016 for all countries in the sample. Thus, finding unit root problems for those variables was unsurprising. To counteract the unit root problems, first differences are used for the variables 𝑇𝐸𝑇𝐴𝑖,𝑡, 𝐶𝐴𝑅𝑖,𝑡 and 𝐷𝐶𝐺𝐷𝑃𝑐,𝑡 from this point onwards. Also, due to the high correlation between 𝑃𝐵𝑇𝑇𝐴𝑖,𝑡−1 and 𝑂𝑅𝑂𝐸𝑖,𝑡 the main focus of the empirical analysis will include only 𝑃𝐵𝑇𝑇𝐴𝑖,𝑡−1, regressions including 𝑂𝑅𝑂𝐸𝑖,𝑡 will be explored in the robustness checks. Delis and Kouretas (2011) include both the lagged variable of 𝑃𝐵𝑇𝑇𝐴𝑖,𝑡 and total revenue to total expenses. However, in my opinion, due to the high correlation, that would empirically be identical to including either variable and its corresponding lag. Therefore, the method of including both 𝑂𝑅𝑂𝐸𝑖,𝑡 and the lagged variable of 𝑃𝐵𝑇𝑇𝐴𝑖,𝑡 is abandoned in the main analysis. A robustness check will provide details on this matter.
19 As Delis and Kouretas (2011) point out in their paper, there is a reverse causation problem between the dependent variable 𝑅𝑊𝐴𝑇𝐴𝑖,𝑡 and the bank control variables 𝑇𝐸𝑇𝐴𝑖,𝑡, 𝑃𝐵𝑇𝑇𝐴𝑖,𝑡−1 and 𝑂𝑅𝑂𝐸𝑖,𝑡. The paper aims to estimate the effect of the Taylor rule residual as a proxy for interest rates on the ratio risk-weighted assets to total assets as a proxy of bank risk-taking. However, bank risk-taking also affects the interest rate. The ECB its main goal is to maintain stable prices in order to increase economic welfare and the growth potential of an economy. Bank risk-taking leads to a deterioration of economic welfare and therefore influences the interest rates. To counter this endogeneity problem, I will include one period lagged variables for 𝑇𝐸𝑇𝐴𝑖,𝑡, 𝑃𝐵𝑇𝑇𝐴𝑖,𝑡−1 and 𝑂𝑅𝑂𝐸𝑖,𝑡 as instruments.
The fixed effects model in Table 3 column I shows insignificant negative results for Taylor rule residual. Therefore, I cannot conclude anything about the relationship between the Taylor rule residual and bank risk-taking based on this specification of the model. Significant results are found for the interaction term, GDP growth, inflation, lagged ratio profits before tax to total assets, logarithm of total assets, domestic credit compared to GDP, ratio quantitative easing to GDP, and the constant. These results are in line with the hypotheses, except for quantitative easing to GDP and the interaction term. A possible explanation for the negative relationship of the effect of quantitative easing might be the lack of data. The quantitative easing program started in October 2014, which gives this research lack of input to measure the effect. Another possible explanation can be that when the interest rate is approaching zero, QE loses its effectiveness. As for the interaction term, possible explanations for the positive effects could result from the fact that the results towards the interaction term could be biased as 𝐶𝐴𝑅𝑖,𝑡 contains a unit root problem. Also, inflation has a positive relationship to bank risk-taking, thus finding evidence in favour of the interest margin hypothesis described above.
20 To approach the method implemented by Delis & Kouretas (2011), in column III, both 𝑃𝐵𝑇𝑇𝐴𝑖,𝑡−1 and 𝑂𝑅𝑂𝐸𝑖,𝑡 are included. As can be observed from Table 3, logarithm of total assets, domestic credit compared to GDP and the ratio quantitative easing to GDP are significant. All have the same sign as in columns I and II. This confirms the relationships found before for these control variables. However, this model does reject the Hausman test meaning that the instrumental variable specification in column III might not be suitable for the model. In general, the Hausman analyses the difference between two estimators from an OLS regression and an IV regression. The null hypothesis reads that both estimators are consistent. The alternative hypothesis reads that only the instrumental variables regression is consistent. The tests saves the residuals from the first stage regression and adds them to the model of interest. Concluding that the results in column III have to be interpreted with caution.
A fourth test is executed by extending the model with data on Main Refinancing Operations (MRO) from the ECB as after the best-practises of Heins (2017). The model has been run with 𝑀𝑅𝑂𝑡 included in the control variables. 𝑀𝑅𝑂𝑡 is significant and positive for different model specifications, as column IV and V show, IV for including 𝑃𝐵𝑇𝑇𝐴𝑖,𝑡−1 and V for including 𝑂𝑅𝑂𝐸𝑖,𝑡. Concluding that MRO positively influence the risk-taking of banks. Which is against expectations while interest rates are assumed to negatively influence risk-taking. However, these results might be due to the endogeneity of MRO. Also, when including the 𝑀𝑅𝑂𝑡, for both specifications of the model, the Taylor rule residual is significant with the expected sign. The rest of the different control variables which are significant in either column IV and/or column VI have the same sign as in the benchmark model projected in column I. However, the Hausman test does strongly reject the specification of the 2SLS within estimator instrumental variable model in column V. Thus the results in column V have to be interpreted with caution.
21 World Bank, Bank Regulation and Supervision Survey. I expect the regulatory indexes to have a negative relationship to taking. This because if regulations increase, I expect bank risk-taking to decrease. Delis and Kouretas (2011) mention that not including these variables will lead to a serious omitted variables bias. This because countries in the Eurozone are highly subjective to regulatory, macroeconomic and structural conditions. As mentioned before, I do not include the indexes because of their limited availability. As a robustness check, column VI shows the results of the Equation (1) including the three indexes. Note, Delis and Kouretas (2011) mention that the endogeneity problem described above might also exist with respect to the regulatory indexes. However, due to the limited amount of data available, I will not include instrumental variables with respect to the regulatory indexes. The results in column VI show that none of the variables are significant anymore after including the indexes. Also, the Hausman test strongly rejects this specification of the 2SLS within estimator instrumental variable model. Thus the results have to be interpreted with caution.
22
Table 3: Regressions 2SLS within estimator instrumental variable regressions
I II III IV V VI VII TRR -0.006 -0.012* -0.001 -0.014* -0.018** -0.088 -0.007 (0.007) (0.007) (0.017) (0.008) (0.008) (0.363) (0.006) FDCAR 0.003 -0.004 0.007 0.003 -0.003 -0.024 0.003 (0.003) (0.004) (0.013) (0.003) (0.004) (0.067) (0.003) ircr 0.002*** 0.002*** 0.001 0.002*** 0.002*** 0.005 0.002*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.015) (0.000) FDTETA -0.681 0.020 -1.131 -0.746 -0.058 25.191 -0.716 (0.608) (0.662) (1.302) (0.604) (0.662) (63.355) (0.447) ggdp 0.004* 0.000 0.006 -0.001 -0.003 -0.006 0.003** (0.002) (0.002) (0.009) (0.003) (0.003) (0.056) (0.002) infl 0.009* 0.013*** 0.009 -0.003 0.003 -0.122 0.010*** (0.005) (0.004) (0.006) (0.007) (0.006) (0.507) (0.004) lagPBTTA 2.358** 3.279 2.145** -4.933 2.223*** (1.064) (4.611) (1.049) (22.942) (0.804) LOGTA -0.101*** -0.067*** -0.116* -0.104*** -0.071*** 0.118 -0.103*** (0.014) (0.016) (0.069) (0.014) (0.016) (0.605) (0.012) FDDCGDP 0.001* 0.001** 0.001* 0.001 0.001** -0.001 0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.007) (0.000) QEGDP -0.003** -0.002* -0.003* -0.002* -0.001 -0.002*** (0.001) (0.001) (0.002) (0.001) (0.001) (0.001) OROE 0.222*** -0.137 0.214*** (0.053) (0.506) (0.053) MRO 0.016** 0.013** (0.006) (0.006) caprq 0.005 (0.049) mdisc 0.068 (0.319) offpr 0.057 (0.176) Constant 2.361*** 1.503*** 2.780 2.401*** 1.571*** -1.895 2.368*** (0.255) (0.317) (1.771) (0.254) (0.318) (11.855) (0.211) Hausman 0.003 0.083 0.161 0.007 0.115 1.000 0.000 Observations 737 727 727 737 727 115 718 Number of ID 87 86 86 87 86 67 82
Notes: The table reports the coefficients and robust standard errors in parentheses. The dependent variable is the
ratio of risk-weighted assets to total assets. Columns I-VII show two-stage least-squares within estimator instrumental variable regressions. The variables are: TRR is the Taylor rule residual, FDCAR is the first difference capital adequacy ratio – tier 1, ircr is the interaction variable between Taylor rule residual and capital adequacy ratio – tier 1, FDTETA is the first difference ratio of total equity to total assets, ggdp is the growth rate of GDP, infl is the inflation rate, lagPBTTA is the lagged ratio of profits before tax to total assets, LOGTA is the natural logarithm of total assets, FDDCGDP is the first difference domestic credit provided by the banking sector as a share of GDP, QEGDP is the ratio of quantitative easing to GDP for the Eurozone, OROE is the ratio of operating revenue to operating expenses, MRO is the main refinancing operations rate of the ECB, caprq is the index of capital requirements, offpr is the supervisory power index, mdisc is the market discipline index. The Hausman test reports results from the Hausman test testing the fixed effects regressions against the instrumental variable regressions.
23
6. Conclusions
The paper has investigated the relationship between the low interest rate environment and bank risk-taking. For the empirical results, an unbalanced panel dataset of 106 banks within the Eurozone for the period 2002-2016 is used. A fixed effects model corrected for autocorrelation and heteroskedasticity finds weak negative relationship between the Taylor rule residual and the ratio risk-weighted assets to total assets in a few specifications of the model. Other results are that the quantitative easing program of the ECB lowers risk-taking behaviour of banks, which is robust to most specifications of the model. The results of quantitative easing to aggregate GDP are not in line with the hypothesis. The hypothesis predicted a positive relationship. As explained, this could be due to the limitation of data, loss of effectiveness of QE due to low interest rates and the increase of lending of banks. The Main Refinancing Operations rate positively impacts bank risk-taking, which is not in line with the hypothesis. Explanations could be found in the endogeneity of the variable.
24 The paper is limited in certain aspects. Some variables, as indicated before, are omitted due to either collinearity or data limitations. Data was limited in terms of the amount of banks, the completeness of the dataset and the amount of variables available. The number of banks in the sample could be expanded by including non-publicly traded banks. Including more, possibly smaller, banks might lead to different conclusions. An expanded and more balanced dataset should lead to results which reflect reality even better. Another suggestion would be to find different measures for the regulatory indexes, while these are limited in availability. Lastly, the effect of omitting some variables, due to data limitations, might have impacted the results. Thus the results presented should be read with caution. I encourage future research to expand the research conducted above.
Next to recommendations made before, future research could focus on the effects of the monetary policy on financial stability. The literature has recommended studying this relationship, but direct measures of financial stability have yet to be found. Next to this, future research could investigate the effect of quantitative easing on policymaking while data on this is still limited due to the novelty of the program. A recommendation to estimate this would be to use more frequent data. Next to using more frequent data, it would be interesting to replicate this study with data from the United States and Japan. These countries also had a quantitative easing program in place.
25
7. Reference list
Adesina, K. & Mwamba, J., 2016. Do Basel III Higher Common Equity Capital Requirements Matter for Bank Risk-taking Behaviour? Lessons from South Africa. African Development Review, 28(3), pp. 319-331.
Apel, M. & Claussen, C. A., 2012. Monetary policy, interest rates and risk-taking. Sveriges Riksbank Economic Review, Volume 2, pp. 68-84.
Basel Committee on Banking Supervision, 2011. Basel III: A global regulatory framework for more resilient banks and banking systems. Bank for International Settlements, pp. 1-7
Bikker, J. A. & Vervliet, T. M., 2017. Bank Profitability and Risk-Taking under Low Interest Rates. Utrecht University School of Economics Tjalling C. Koopmans Research Institute, Discussion Paper Series 17-10, pp. 1-38.
Borio, C. & Zhu, H., 2012. Capital regulation, risk-taking and monetary policy: A missing link in the transmission mechanism?. Journal of Financial Stability, Volume 8, pp. 236-251.
Chen, M., Jeon, B. N. & Wang, R., 2017. Monetary policy and bank risk-taking: Evidence from emerging economies. Emerging Markets Review, Volume 31, pp. 116-140.
Cociuba, S. E., Shukayev, M. & Ueberfeldt, A., 2016. Collateralized borrowing and risk-taking at low interest rates. European Economic Review, Volume 85, pp. 62-83.
Cœuré, B., 2016. Assessing the implications of negative interest rates. European Central Bank. https://www.ecb.europa.eu/press/key/date/2016/html/sp160728.en.html [Accessed 3 January 2018].
26 Delis, M. D. and Kouretas, G. P., 2011. Interest rates and bank risk-taking. Journal of Banking & Finance, Volume 35, pp. 840-855.
Dell’Ariccia, G., Laeven, L. & Suarez, G. A., 2017. Bank Leverage and Monetary Policy’s Risk-Taking Channel: Evidence from the United States. The Journal of Finance, 72(2), pp. 613-654.
Dell'Ariccia, G., Laeven, L. & Marquez, R., 2013. Real interest rates, leverage, and bank risk-taking. Journal of Economic Theory, Volume 149, pp. 65-99.
Diez de los Rios, A. & Shamloo, M., 2017. Quantitative Easing and Long-Term Yields in Small Open Economies. International Monetary Fund, WP/17/212, pp. 1-45.
Dubecq, S., Mojon, B. & Ragot, X., 2009. Fuzzy Capital Requirements, Risk-Shifting and the Risk Taking Channel of Monetary Policy. Banque de France, Volume 254, pp. 1-43.
European Central Bank., 2017. Asset purchase programmes.
https://www.ecb.europa.eu/mopo/implement/omt/html/index.en.html [Accessed 3 January 2018].
Heins, R., 2017. Taylor Rule Residual and Bank Risk-Taking. University of Groningen, pp. 2-37.
Hull, J. C., 2015. In: Risk Management and Financial Institutions. Hoboken, New Jersey: John Wiley & Sons, p. 338.
Inman, P., 2016. Irish economy surges 26% as revised figures take in foreign investment. https://www.theguardian.com/business/2016/jul/12/irish-economic-growth-revised-figures-foreign-investment-aircraft [Accessed 4 January 2018].
Jones, C., 2017. Financial Times: "Central banks warned on risk from low inflation".
27 Maddaloni, A. & Peydró, J.-L., 2008. Bank Risk-taking, Securitization, Supervision, and Low Interest Rates: Evidence from the Euro-area and the U.S. Lending Standards. The Review of Financial Studies, 24(6), pp. 2121-2165.
Meckling, W. H. & Jensen, M. C., 1976. Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure. Journal of Financial Economics, 3(4), pp. 305-360.
Myers, S. C., 1977. Determinants of Corporate Borrowing. Journal of Financial Economics, Volume 5, pp. 147-175.
Nucera, F., Lucas, A., Schaumburg, J. & Schwaab, B., 2017. Do negative interest rates make banks less safe?. European Central Bank, pp. 1-17.
Oliveira de Moraes, C., Montes, G. C. & Pereira Antunes, J. A., 2016. How does capital regulation react to monetary policy? New evidence on the risk-taking channel. Economic Modelling, Volume 56, pp. 177-186.
Özşuca, E. A. & Akbostancı, E., 2016. An Empirical Analysis of the Risk-Taking Channel of Monetary Policy in Turkey. Emerging Markets Finance & Trade, Volume 52, pp. 589-609.
Paligorova, T. & Santos, J., 2017. Monetary policy and bank risk-taking: Evidence from the corporate loan market. J. Finan. Intermediation, Volume 30, pp. 35-49.
Rajan, R. G., 2006. Has finance made the world riskier?. European Financial Management, Volume 12, pp. 499-533.
Reuters., 2018. ECB should fix date to end bond purchases: Weidmann. https://www.reuters.com/article/us-ecb-policy-weidmann/ecb-should-fix-date-to-end-bond-purchases-weidmann-idUSKBN1EW0HE?il=0 [Accessed 18 January 2018].
29
8. Appendixes
Appendix A – List of Countries in Eurozone and year of entry
Table 4: Eurozone countries and year of entry
Country Year of entry
30
Appendix B – Variable description
Table 5: Variable descriptions
Variable Description Source
EONIA The EONIA is calculated as the weighted rate for the
overnight maturity, calculated by collecting data on unsecured overnight lending in the euro area provided by banks
belonging to the EONIA panel. The data is on a daily average level. To proceed to a yearly level, the average of all days is taken.
ECB data warehouse
Inflation Rate The data is available on an annual average rate of change (%). This is the average of the percentage change on the same period of the previous year. The inflation rate is constructed by comparing the price of a basket of goods over time.
Eurostat database
Chained GDP level The data is the unadjusted GDP at chained linked volumes obtained from the Eurostat. It is available in annual time series. Eurostat database calculates this measure in million Euros. The measure is obtained by applying previous year's price's growth rates to the current price figure of a specific reference year, 2010.
Eurostat database
Current GDP level The data is the unadjusted GDP at current prices. It is available in annual time series.
Eurostat database
Potential GDP level This measure is obtained from Economic Outlook No 101 - June 2017 by OECD statistics. It is defined as the level of output that an economy can produce at a constant inflation rate, thus potential real GDP. The measure shows the value of potential output of total economy per country on an annual basis.
OECD statistics
Quantitative Easing The measure shows the history of cumulative purchases under the Asset Purchasing Programme of the ECB. The data shows the history of cumulative purchases under the APP end of the month. Thus the last month of the year is included as an annual measurement.
European Central Bank
Capital adequacy ratio - Tier 1
Calculated as the ratio of Tier 1 Capital to total risk-weighted assets, in accordance with banking regulations and expressed as a percentage. Tier 1 Capital includes common shareholders’ equity and qualifying preferred stock, less goodwill and other adjustments.
Datastream
Total capital Sums common equity, preferred stock, minority interest, long-term debt, non-equity reserves and deferred tax liability in untaxed reserves.
Datastream
Risk-weighted assets Contains the total of the carrying value of each asset class multiplied by their assigned risk weighting, as defined by banking regulations.
31
Table 5: Variable descriptions (Continued)
Variable Description Source
Total assets Sums cash & due from banks, total investments, net loans, customer liability on acceptances (if included in total assets), investment in unconsolidated subsidiaries, real estate assets, net property, plant and equipment and other assets.
Datastream
Profits before tax Estimated as Operating Income: Interest Income + Total Non-Interest Income – Non-Interest Expense-Total – Non- Non-Interest Expense – Provision for Loan Losses.
Datastream
Operating revenue Represents the total operating revenue of the bank. Datastream Operating expense Sum of all expenses related to operations. Datastream Importance of banks is
the domestic credit provided by the banking sector as a share of GDP
Measured as a percentage of GDP. World Bank
Main refinancing operations
Measured in percentages. European Central
32
Appendix C – Descriptive statistics Table 6: Descriptive statistics per country
AT BE EE FI FR DE GR IE IT LU NL PT SK SI ES RWATA 0.604 0.306 0.491 0.406 0.419 0.427 0.646 0.582 0.625 0.104 0.458 0.673 0.624 0.584 TRR (%) 1.315 1.366 3.315 0.962 0.857 0.703 1.632 0.746 1.285 1.860 1.007 1.307 1.625 1.793 1.658 CAR (%) 9.870 12.362 16.055 10.502 9.416 10.429 11.607 10.817 10.171 8.171 12.998 9.291 13.956 8.566 9.928 TETA 0.064 0.313 0.084 0.056 0.096 0.066 0.081 0.059 0.105 0.012 0.053 0.049 0.087 0.068 0.073 QEGDP (%) 1.392 1.392 3.479 1.392 1.392 1.392 1.392 1.392 1.392 1.392 1.392 1.392 2.609 2.087 1.392 ggdp (%) 1.419 1.435 3.412 1.116 1.089 1.195 -0.352 4.652 -0.023 2.837 1.162 0.190 1.939 0.892 1.394 infl (%) 1.860 2.029 2.317 1.627 1.553 1.453 2.214 1.593 1.840 2.562 1.653 1.853 1.250 1.840 2.087 PBTTA 0.005 0.010 0.004 0.007 0.010 0.001 -0.008 0.003 0.008 0.004 0.007 0.003 0.007 -0.005 0.011 LOGTA 16.691 18.035 13.081 15.031 17.203 16.394 16.771 18.355 16.876 18.032 18.862 17.472 15.362 15.369 18.818 OROE 1.126 5.305 1.155 1.238 1.217 1.096 0.989 1.093 1.122 1.075 1.109 1.072 1.259 0.982 2.317 DCGDP (%) 127.640 117.746 76.548 126.421 132.599 138.018 121.188 173.451 140.260 168.593 191.824 169.349 67.243 83.585 195.317
Notes: The table reports the descriptive statistics per country for the variables used in the regression analysis.
Country codes are: AT is Austria, BE is Belgium, EE is Estonia, FI is Finland, FR is France, DE is Germany, GR is Greece, IE is Ireland, IT is Italy, LU is Luxembourg, NL is the Netherlands, PT is Portugal, SK is Slovakia, SI is Slovenia and ES is Spain.
33
Appendix D – Graph RWATA and TRR
Figure 1: Scatterplot RWATA on TRR including outliers
Figure 2: Scatterplot RWATA on TRR excluding outliers
34
Appendix E – Robustness checks
Table 7: Fixed effects regressions
I II III IV V VI VII TRR -0.002 -0.005 -0.006 -0.011 -0.012* 0.030 -0.003 (0.006) (0.006) (0.006) (0.007) (0.007) (0.019) (0.006) FDCAR -0.004*** -0.005*** -0.005*** -0.005*** -0.005*** -0.005** -0.004*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.002) (0.001) ircr 0.001*** 0.002*** 0.002*** 0.002*** 0.002*** 0.001 0.001*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) FDTETA 0.472* 0.430** 0.494** 0.422** 0.489** 1.817*** 0.448* (0.253) (0.185) (0.200) (0.183) (0.197) (0.662) (0.253) ggdp 0.005*** 0.003 0.003 -0.000 -0.001 0.007 0.005*** (0.002) (0.002) (0.002) (0.003) (0.003) (0.006) (0.001) infl 0.016*** 0.017*** 0.016*** 0.009 0.006 0.061** 0.016*** (0.004) (0.003) (0.003) (0.006) (0.006) (0.025) (0.004) lagPBTTA 1.153** 0.712** 0.752** -0.096 1.228** (0.498) (0.335) (0.315) (0.704) (0.563) LOGTA -0.082*** -0.068*** -0.070*** -0.072*** -0.075*** -0.102*** -0.083*** (0.016) (0.017) (0.017) (0.017) (0.017) (0.018) (0.016) FDDCGDP 0.001** 0.001** 0.001** 0.001** 0.001* 0.002*** 0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.001) (0.000) QEGDP -0.002** -0.002** -0.002** -0.001 -0.002* -0.002** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) OROE 0.110*** 0.095** 0.109*** 0.092** (0.035) (0.036) (0.035) (0.036) MRO 0.012* 0.012* (0.007) (0.007) caprq -0.003 (0.004) mdisc -0.024 (0.015) offpr -0.011* (0.006) Constant 2.003*** 1.630*** 1.693*** 1.693*** 1.765*** 2.434*** 2.006*** (0.284) (0.314) (0.314) (0.316) (0.316) (0.345) (0.282) Observations 788 779 779 779 779 161 769 R-squared 0.312 0.318 0.323 0.322 0.327 0.649 0.407 Number of ID 87 86 86 86 86 71 82
Notes: The table reports the coefficients and robust standard errors in parentheses. The dependent variable is the
ratio of risk-weighted assets to total assets. Columns I-VII show fixed effects regressions corrected for heteroskedasticity and autocorrelation. The variables are: TRR is the Taylor rule residual FDCAR is the first difference capital adequacy ratio, ircr is the interaction variable between Taylor rule residual and capital adequacy ratio – tier 1, FDTETA is the first difference ratio of total equity to total assets, ggdp is the growth rate of GDP, infl is the inflation rate, lagPBTTA is the lagged ratio of profits before tax to total assets, LOGTA is the natural logarithm of total assets, FDDCGDP is the first difference domestic credit provided by the banking sector as a share of GDP, QEGDP is the ratio of quantitative easing to aggregate GDP for the Eurozone, OROE is the ratio of operating revenue to operating expenses, MRO is the main refinancing operations rate of the ECB, caprq is the index of capital requirements, mdisc is the market discipline index and offpr is the supervisory power index. *** Statistical significance at the 1% level.
35
Appendix F – Composition of Indexes
For the composition of the indexes, the following questions have been analysed. The database used is from the World Bank, Bank Regulation and Supervision Survey. The questionnaire has data for 2003, 2007 and 2012. For each question, answering ‘yes’ gives the value 1 and answering ‘no’ gives the value 0. All values are added or subtracted. The maximum score for the capital requirement index is 6, the maximum score for the market discipline index is 7 and the maximum score for the supervisory power index is 14.
Table 8: Composition indexes
Capital Requirement Index
1.5 Are the sources of funds to be used as capital verified by the regulatory/supervisory authorities?
1.6 Can the initial disbursement or subsequent injections of capital be done with assets other than cash or government securities? 1.7 Can initial disbursement of capital be done with borrowed funds? 3.1.1 Is this ratio risk-weighted in line with the Basle guidelines? 3.3 Does the minimum ratio vary as a function of market risk? 3.9.1 3.9 Before minimum capital adequacy is determined, which of the
following are deducted from the book value of capital?
Market value of loan losses not realized in accounting books?
3.9.2 Unrealized losses in
securities portfolios?
3.9.3 Unrealized foreign
exchange losses?
Market Discipline Index
3.5 Is subordinated debt allowable as part of capital? 5.1 Is an external audit a compulsory obligation for banks?
8.1 Is there an explicit deposit insurance protection system? If no, you may skip to question 8.2. If yes:
10.1 Does accrued, though unpaid, interest/principal enter the income statement while the loan is still performing?
10.3 Are financial institutions required to produce consolidated accounts covering all bank and any non-bank financial subsidiaries?
10.4.1 Are off-balance sheet items disclosed to the public?
10.5 Must banks disclose their risk management procedures to the public? 10.6 Are bank directors legally liable if information disclosed is
erroneous or misleading?
10.7 Do regulations require credit ratings for commercial banks?
Supervisory Power Index
5.5 Does the supervisory agency have the right to meet with external auditors to discuss their report without the approval of the bank? 5.6 Are auditors required by law to communicate directly to the
supervisory agency any presumed involvement of bank directors or senior managers in illicit activities, fraud, or insider abuse? 5.7 Can supervisors take legal action against external auditors for
36
Table 8: Composition indexes (Continued)
Supervisory Power Index
6.1 Can the supervisory authority force a bank to change its internal organizational structure?
10.4 Are off-balance sheet items disclosed to supervisors? 11.2 Can the supervisory agency order the bank's directors or
management to constitute provisions to cover actual or potential losses?
11.3.1 Can the supervisory agency suspend the directors' decision to distribute:
Dividends?
11.3.2 Bonuses?
11.3.3 Management fees?
11.6.1 Who can legally declare - such that this declaration supersedes some of the rights of shareholders - that a bank is insolvent:
Bank supervisor
11.7.1 According to the Banking Law, who has authority to intervene - that is, suspend some or all ownership rights - a problem bank?
Bank supervisor
11.9.1.1 Regarding bank restructuring and reorganization, can the supervisory agency or any other government agency do the following:
Supersede shareholder rights?
Bank supervisor 11.9.1.3 Deposit insurance agency 11.9.1.4 Bank restructuring or Asset Management Agency
11.9.1.5 Other (please specify)
11.9.2.1 Remove and replace management? Bank supervisor
11.9.2.2 Court 11.9.2.3 Deposit insurance agency 11.9.2.4 Bank restructuring or Asset Management Agency
11.9.2.5 Other (please specify)
11.9.3.1 Remove and replace directors? Bank supervisor
11.9.3.2 Court 11.9.3.3 Deposit insurance agency 11.9.3.4 Bank restructuring or Asset Management Agency
