Bank risk-taking and the slope of the yield curve
J.R. Venema
University of Groningen
Faculty of Economics and Business
Msc. Finance
Supervisor: dr. J.O. Mierau
June 2015
Abstract
One of the basic tasks of banks is to safe keep deposits and to transform these deposits into long-term loans. In this paper I examine the relationship between the risk-taking behaviour and the slope of the yield curve which is used as a measure of the profitability of the strategy. I use three types of risk-measures, namely: risk assets, non-performing loans, and the Z-score which is commonly used in the banking literature. Overall I find significant but mixed results. An increase in the slope of the yield curve causes banks to invest more into the risk free alternatives whereas the amount of non-performing loans increases with an increase in the slope suggesting more risk-taking. Furthermore the Z-score implies that the banks risk taking behaviour also increases with an increase in the slope. Overall the results imply a more risky behaviour of banks with an increase in the slope.
JEL classifications E43, E52, G21
2 1. Introduction
The main reason of the existence of financial intermediaries is because of asymmetric information. Without asymmetric information, the financial markets can find solutions to financing problems of their own. This resulted in the core task of the banks, to safe keep deposits and to transform these deposits into long-term loans. Banks are better than individuals in this maturity transformation because banks are specialized in monitoring and profit from economies of scale. However, depositors can claim their money at any time but if they would all go at the same time, the money is not there because it is transformed into long-term loans. This implies that the banks have to convince their depositors that they follow a sound strategy in order to keep the money. The profitability of this strategy is determined by the difference between the long-term and short-term interest rates, also known as the slope of the yield curve. Traditionally, banks following this strategy have short term deposits and long term investments and this maturity transformation is their core business model. However, the crisis of 2008 has shown that modern banking includes more. In the crisis banks had to little high quality capital, held insufficient liquid funds and were highly levered. The Basel committee recognized this and increased the amount of regulation for banks. More regulation decreases the profitability of the core business model and thereby increasing incentives for risk taking. I would like to present a sample of this occurrence with the help of the bank balance sheet using fictional figures. A bank has a total on the asset side of 100 which are distributed in 90 to loans and 10 to liquid assets, the latter is required to provide the depositors cash and because of regulation for capital buffers. The liquid assets are not producing any income for the bank. Furthermore, the bank is funded with 80 by depositors and 20 by the shareholders. All else constant, an increase in the capital buffer to 12 results in a reduction of the loans by 2. The loss of interest revenue is borne by the shareholders, and obviously they are not happy about it. To keep the interest revenue constant banks can take on more risky loans. In the real world there is another problem, interest rates do not stay constant and therefore the increase in bank risk-taking behaviour can be seen when the difference between the long- and short-term interest rate decreases (a decrease in the slope of the yield curve). But is there a change in the risk-taking appetite by banks following a change in the slope of the yield curve?
Banks searching for return is a phenomenon mentioned in the literature. In the paper of Rajan (2005) he states that both the bank itself and the investment manager can have the incentive to increase their risk-taking behaviour as a result of searching for return. From the banks’ perspective, if their long-term fixed interest rate liabilities are relatively high, a change in the interest rate environment from high to low interest rates induces banks to increase their risk taking behaviour. From the investment managers’ perspective, nowadays compensation relies heavily on returns, and clients are attracted by good returns. This implies that good returns are important for investment managers thereby raising the incentive towards more risk-taking by investment managers. Therefore giving more fuel to the question, how does the slope of the yield interact with bank risk-taking?
Ever since the crisis of 2008 there is a close eye on bank risk-taking. Furthermore, from the perspective of Sibbertsen, Wegener & Basse (2014) there is a break in the yield spreads of EMU government bonds. Clearly, looking at the interest rates and slope of the yield curve development (figure 1.) something changed in 2008. It would make sense that after the crisis of 2008 banks are more cautious with respect to risk-taking which raises the question; did banks change their behavior with respect to risk-taking?
To research the slope of the yield curve in relation to bank risk-taking, I employ three measures of risk namely; risk-assets, non-performing loans, and the Z-score. The risk assets measurement is based on the volatility of the banks their portfolio, the non-performing loans measures the quality of loans given by banks, and the Z-score indicates the probability of default of the bank. I create a large unbalanced panel data set from banks in Europe, given that the country has adopted the euro, in the period of 2001-2013.
The results of the main analysis suggest that banks decrease their risks assets as a result of an increase in the slope of the yield curve. Furthermore, an increase in the slope also causes a higher amount of non-performing loans. Also, the Z-score indicates that there is a higher probability of default given a rise in the slope. These results suggest that, although the banks are increasing their investment into risk free alternatives, the risk-taking behaviour of banks increases with an increase in the slope. I also test if the incentive is coming from the banks which are searching for return. In line with the counter argument provided by De Nicolò et al. (2010), less capitalized banks are more likely to engage in risky behaviour. However, the results do not support this as banks which are more capitalized are more likely to increase their risk-taking behaviour.
4 result of the worse conditions after the crisis and thus the slope of the yield curve did not contribute to that change. The change in the Z-score is in line with the survivorship bias of the sample. It is therefore hard to make conclusions other than that banks are more cautious.
The remainder of this thesis is structured as follows. Section 2 provides with literature related to the topic and derives the hypothesis. Section 3 describes the data and methodology. In Section 4 I conduct the empirical analysis and discuss the results. Finally, Section 5 concludes.
2. Literature review and hypothesis development 2.1 Usefulness of the yield curve
The usefulness of the yield curve started with the article Estrella & Mishkin (1998) who examined the performance of various financial variables in predicting future U.S. recessions. This is confirmed by Ozturk & Pereira (2013) who state that the yield curve is a very important measurement in economics as it shows how economic activity will develop in the future. Furthermore, Beck, Jakubik & Piloiu (2013) are stating that that a drop in global economic activity remains the most important risk for bank asset quality. The yield curve describes the relationship between the short- and long-term interest rates. In general the yield curve has an upward slope because the short-term interest rates are lower than the long-term interest rates. Ozturk & Pereira (2013) question the yield curve as a predictor and develop their hypothesis based on three arguments. The fact is that long-term interest rates are reflections of the expected future short-term interest rates. A countercyclical monetary policy can induce this reduction in interest rate expectation. When there is a monetary policy tightening, short-term interest rates rise but the long-term interest rates do not raise as much which implies that the slope of the yield-curve decreases. The last argument is that consumers prefer to have a stable level of income and thus “flight to quality” in more volatile circumstances. Ozturk & Pereira (2013) find that the yield curve can be used to predict recessions one year ahead.
2.2 In search for return
Not only the institution can have the incentive to search for return, investment managers also have an incentive to take more risk. One of the arguments Rajan (2005) makes in his article is that in the deregulated, competitive environment, investment managers cannot be provided the same staid incentives as banks of yore. In the new environment managers have the incentive to search for good investments; this is because their compensation relies on good returns. Also, investors influence the incentive of the investment managers because new investors are attracted by high returns. As stated by Rajan (2005), compensation is typically related to the amount of assets under management. If the investment managers are not delivering up to expectations, the investors can leave but in general they don’t leave fast. This enhances the incentive to take more risks even more.
2.3 The interest rate environment
One way the incentive to search for more return can be triggered is via a change in the interest rate environment which can be caused by monetary policy. In the short run, monetary policy can change the short-term interest rate. For monetary policy to change the long-term interest rate the policy has to be aimed at changing the future expectations of the short-term interest rate. One way monetary policy can influence the interest rate is by monetary expansion. Buch, Eickmeier & Prieto (2014) state that once expansionary monetary policy is employed, the interest rate will go down because of the increase in money supply. As a result, banks will soften their lending standards and are going for projects with higher risk so their yield is higher. The increase in higher risk projects will affect the balance sheet of the bank towards more risk-taking, and thus increasing the chance of failure. Buch et al. (2014) build on this idea and question whether or not there is an incentive for bank managers to keep he average returns constant and thus in a low interest rate environment shifting credit supply towards more risky segments. They find evidence that small banks do increase their exposure to risk, but they only provide more credit towards more risky projects and charge a lower risk premium. However, Buch et al. (2014) find that these small banks do change their contractual terms to compensate for the increase in risk.
6 causes banks to increase their leverage and also causes banks to lower their monitoring and thus increasing risk taking. Furthermore, highly capitalized banks are monitoring less when there is a reduction in the risk-free rate, while the opposite is true for poorly capitalized banks. Similar results have been found by Delis & Kouretas (2011) who are also interested in the low interest rate environment. They find that the interest rates have a negative effect on bank risk-taking. The low interest rate environment has a bigger impact on banks that are engaging in modern banking and banks with high level of capitalization.
A counter argument towards bank risk-taking and interest rate is discussed in De Nicolò et al. (2010) that at least in the short run, there is also an opposite risk-shifting effect when financial intermediaries operate with limited liability. Then the balance is dependent on the degree of limited liability and financial health of the financial intermediary. The low interest rates can trigger increased risk taking behavior for well-capitalized banks and poorly capitalized banks decrease their risk taking behavior. Capital requirement would reduce the risk-taking behavior as well. The more “skin-in-the-game” the financial intermediary has, the less it is willing to risk because it has too much to lose in case of failure. The same reasoning goes for banks with a lot of potential to grow which have more to lose than banks close to bankruptcy. One problem which enhances this moral hazard is the safety net as discussed by Dam & Koetter (2012). Because there is a higher bail-out expectations banks are even more willing to take on risk, this effect is even stronger for banks that are “too-big-to-fail”.
Is there a change in the risk-taking appetite by banks following a change in the slope of the yield curve? According to the literature there are quite a few reasons to expect the interest rate (and therefore the slope) to be linked towards the bank risk-taking. Mainly the change in interest rate causes banks to search for return, whereas investment manager is in search for return as part of compensation thereby increasing the risk taking behaviour. To investigate the slope of the yield curve in relation to bank risk-taking empirically, I construct the following hypothesis:
The slope of the yield has an impact on bank risk-taking.
3. Data and Methodology
For the model I build onto the idea of Delis & Kouretas (2011) who employ a model to capture bank risk-taking in relation towards interest rates. Instead of the interest rates I will use the slope of the yield curve. The general equation for analyzing whether or not the slope of the yield curve has an effect on bank risk-taking will be estimated by using:
where the risk variable, r, of bank i at time t is written as a function of the slope of the yield curve, y, of the country in which a particular bank takes their main residence i at time t, a set of common bank-level control variables of the particular bank i at time t, and a few macroeconomic control variables which are applicable to the country in which the banks reside i at time t. Hereafter I will discuss the specifics of these variables but first I will explain the some of the choices beforehand.
Because of the current monetary policy by the ECB I take the sample from Europe of countries which have adopted the Euro. The sample starts from the year 2001 until the year 2013. The analysis is restricted to countries which have adopted the Euro and thus have a common monetary policy. This means that the following countries only enter the sample after they adopted the Euro: Cyprus (2008), Malta (2008), Slovenia (2008), Slovakia (2009), and Estonia (2011). The data related to banks is collected from Bankscope and is on an annual basis. Because I focus on banks that take deposits only the commercial banks, savings banks, and cooperative banks that are operating in the respective countries. Bankscope provides data on the basis of a consolidation code. This makes room for possible mistakes as there could be double counting involved. In line with Duprey & Lé (2015) I select the consolidation codes based at the most disaggregated level to maximize the sample size and to look for bank specific risk-taking1. In this sample I only use banks that are active and thus creating a so called “survivorship bias”. However, if the slope of the yield curve has a significant effect on bank risk-taking then the effect would be understated. That is, assuming that more risk leads to more bankruptcy in general. And thus the survivorship bias should not be a problem.
The panel data set is unbalanced and contains 34875 observations of 2701 unique banks. Table 1 reports the summary statistics for all of the variables which are used in this research. It is to be noted that the outliers are still in the sample as I assume that the effect of the outliers on this dataset does not alter the results significantly. A point to be made is that Germany and Italy have a huge impact on the sample as the amount of observations mainly comes from these two countries with respectively 20.397 and 6.252 observations. Table 2 provides the respective correlation coefficients between the variables used in the analysis. The correlations between the variables are not above acceptable levels.
1
8 Table 1.
Descriptive statistics
Variables Mean Standard deviation Min Max
Dependent Risk assets 93.276 8.101 3.356 100.000 Non-performing loans 2.964 3.384 0.000 59.814 Z-score -6.407 6.977 -187.860 24.577 Independent Slope 1.418 1.174 -0.650 21.930 Size 13.131 1.446 5.775 21.354 Capitalization 8.706 8.914 -45.817 100.000 Profitability 0.572 2.120 -86.508 193.305 Efficiency 69.298 23.670 0.000 833.289
Off-balance sheet items 10.209 1.755 -0.223 19.565
GDP Growth 0.936 2.386 -8.864 8.276
Domestic credit 108.447 21.543 53.927 305.087
This table reports the summary statistics for the variables that are used in the empirical analysis. The ratios are reported as percentages. The variables used are as follows: the risk assets is the ratio of risky assets to total assets, the non-performing loans is the ratio of impaired loans to gross loans, the Z-score is a measure of a bank’s probability of insolvency, the slope is the long-term interest rate minus the short-term interest rate, size is the logarithm of the total assets, capitalization is the ratio of equity to total assets, profitability is the ratio of profit before tax to total assets, efficiency is the ratio of total revenue to total expenses, off-balance sheet items is the logarithm of off-balance sheet items, GDP growth is the economic growth of the respective country, and Domestic credit is the amount of credit supplied by the financial institutions in the respective country.
3.1 Risk measures
In total I will use 3 risk measures to relate bank risk-taking behaviour towards the slope of the yield curve. Following Delis & Kouretas (2011) I will include the ratio of risk assets to total assets (indicated as Risk assets), and use the ratio of non-performing loans to total loans (indicated as Non-performing loans). The risk assets are the total assets minus the risk-free assets which are the government securities and the cash and balances due from banks. Arguably, the risk asset measure is a direct measure of bank risk-taking as it behaves like the portfolio of the respective bank. An increase in the risk assets implies that banks are more risk-taking because they have less risk-free assets.
Table 2
Correlation matrix
This table reports the correlations for the variables that are used in the empirical analysis. The variables used are as follows: the risk assets is the ratio of risky assets to total assets, the non-performing loans is the ratio of impaired loans to gross loans, the Z-score is a measure of a bank’s probability of insolvency, the slope is the long-term interest rate minus the short-term interest rate, size is the logarithm of the total assets, capitalization is the ratio of equity to total assets, profitability is the ratio of profit before tax to total assets, efficiency is the ratio of total revenue to total expenses, off-balance sheet items is logarithm of off-balance sheet items, GDP growth is the economic growth of the respective country, and Domestic credit is the amount of credit supplied by the financial institutions in the respective country.
Risk assets
Non-performing
loans Z-score Slope Size Capitalization Profitability Efficiency
Off-Balance sheet items GDP Growth Domestic credit Risk assets 1.000 Non-performing loans -0.362 1.000 Z-score 0.155 -0.076 1.000 Slope -0.352 0.251 0.128 1.000 Size 0.096 -0.004 0.281 0.012 1.000 Capitalization -0.253 0.093 -0.829 0.109 -0.307 1.000 Profitability -0.009 -0.063 -0.109 -0.042 -0.023 0.105 1.000 Efficiency 0.01 0.033 -0.013 -0.014 -0.154 0.019 -0.287 1.000
Off-Balance sheet items 0.136 -0.01 0.16 0.023 0.865 -0.174 0.01 -0.175 1.000
GDP Growth 0.17 -0.081 -0.04 -0.413 -0.007 -0.028 0.031 -0.023 0.012 1.000
Domestic credit -0.094 0.175 -0.015 0.263 0.075 0.083 -0.017 -0.043 0.112 -0.117 1.000
is known from the literature that the slope of the yield curve works through credit expansion and contraction.
In addition to the previous measures of risk I will include the Z-score which can be traced back to Boyd & Graham (1986). Lepetit & Strobel (2013) define the Z-score as a risk measure which is commonly used in the empirical banking literature to reflect a bank’s probability of insolvency. Furthermore, Lepetit & Strobel (2013) state that insolvency is a state when the capital-asset ratio and the return on assets are smaller than 0, where car is the capital-asset ratio and roa the return on assets. Lepetit & Strobel (2013) question how to construct this Z-score and propose an alternative approach to construct the Z-score to the one which has been used in previous literature. They find that their proposed way to construct this measure is more straightforward to implement in the assessment of individual bank insolvency and financial stability in general. The general formula for the Z-score is as follows:
The question lies within the calculation of the mean and standard deviation of the return on assets and here I follow the approach of Lepetit & Strobel (2013). Their approach is to use the mean and standard deviation estimates and that are calculated over the full sample *1…T+, combined
with the current values of the car at the respective time t. The Z-score can be interpreted as the inversely related to an upper bound of the probability of insolvency , a higher Z-score indicates that a bank incurs less risk and is more stable. In this dataset I have multiplied the Z-score with -1 so a higher Z-score implies more bank risk-taking which is consistent with the two other risk measures.
3.2 Yield curve
The yield curve is constructed by subtracting the short-term interest rates from the long-term interest rates. For the short-term I use the interest rates on the money markets with the 3 months maturity, and for the long-term rate I use the yield on government bonds with a maturity of 10 years. Figure 1 depicts the monthly interest rates and the slope from January 1995 until March 2015. It should be noted that the data used in the analysis is based on yearly data because it is restricted to the annual bank-level data. Also, information about the long-term interest rates is available per country each year used in the analysis, whereas the short-term interest rates are the same for each country.
and this raises the question whether or not banks changed their risk-taking behavior. Therefore, it would be right to include a regression analysis of the sample before and after 2008. Also from the perspective of Sibbertsen, Wegener & Basse (2014) there is a break in the yield spreads of EMU government bonds. They state that a consequence of the crisis the low yield spreads among government bonds issued by the member states of the EMU now seem to be a phenomenon of happier times. This supports a theory that there might be a structural break in the behaviour of banks after the crisis.
Figure 1. Interest rates and the slope. This figure depicts the movement of the long- and short-term interest rates and the related slope of the yield curve. The data for this table is monthly data gathered from Eurostat about the Euro Area. The long-term interest rate is the EMU convergence criterion series, the short-term is the 3 months money market interest rate, and the slope of the yield curve is the difference.
Generally, the spread of a more risky investment is higher than a lower risk investment. However, it should be noted that in this research the spread is used on a bases of government debt. And thus it is questionable whether or not a higher yield in this regard implies that banks are taking more risks or whether they prefer to invest in government bonds because the yield is high compared to more risky investments. That is, assuming the government bonds are in general risk free and the cost of debt is determined by a risk free rate and a default premium.
Because of the use of government bonds as the upper bound which determines the slope, and the inclusion of government securities in the risk assets, the three dependent variables help to determine which effect is stronger in the research. There is a possibility that not all of the results suggest that the slope of the yield curve has a positive or negative effect on bank risk-taking, thereby creating a possibility to contradict each other. Using three dependent variables helps to overcome
-2.00 0.00 2.00 4.00 6.00 8.00 10.00 In p e rc e n tages Year
12 this problem as the result would always be that all results point into the same direction, or two out of three results point in the same direction (assuming significant results).
3.3 Control variables
There are two sets of control variables that will be used in this research, bank-level control and country-level of control. Control variables should be used to distinguish between results of the dependent variable. If two banks have the same slope and a different outcome for the dependent variable there should be some reason for the answer to be different. More control is not always for the best as stated by Angrist & Prischke (2008). In this case I am interested in the slope of the yield curve in relation to the risk variable. If there is a control variable which depends on the slope of the yield curve then the outcome of the experiment is not good. This is the so called “bad control problem”. According to Angrist & Prischke (2008), good control variables are fixed variables at the time the variable of interest was determined.
hidden risks in these off-balance sheet items. If these items are below a certain value, banks have to purchase them back which results in a loss for the bank. Therefore, it is important to take the off-balance sheet items into account.
For the country-level controls I am inspired by Männasoo & Mayes (2009). The countries in which the banks operate are different and therefore needs to be controlled for. The data is collected from the IMF IFS. The country-level controls that will be used are the GDP-growth rate as a control for macroeconomic development in the respective country, and the domestic credit provided by the banks to GDP.
3.4 Panel data regression
The data used for this research is panel data, mainly describing the behaviour of banks both across time and across entities. The general model for regression, the pooled model, is:
However, in this analysis there is a high suspicion that there is unobserved heterogeneity across banks which is captured by in the regression analysis. These individual-specific effects can be either fixed or random, and depend on the correlation with the regressors. If they are correlated, the fixed effects model will be used, else the random effects. In order to determine which model is appropriate I use the Breusch-Pagan Lagrange Multiplier test to distinguish between the pooled model and the individual-specific model. To determine whether or not the individual-specific effects are correlated or not I use the Hausman test. In this research I will use the fixed effects model by default, because there is a high suspicion that there are correlated individual-specific effects. If this is not the case, the results will state it explicitly.
4. Empirical analysis and results
14 Table 3.
Estimating equation (1) using the full sample, limited controls, and fixed effects
(1)
(2)
(3)
Variables
Risk assets
Non-performing loans Z-score
Slope
-0.790***
0.464***
1.568***
(0.030)
(0.021)
(0.027)
GDP Growth
-0.025**
0.007
0.091***
(0.012)
(0.012)
(0.012)
Domestic credit
-0.022***
-0.016***
0.012***
(0.003)
(0.003)
(0.003)
Constant
96.826***
3.761***
-10.034***
(0.361)
(0.296)
(0.313)
Observations
22,063
7,456
27,489
R-squared
0.055
0.109
0.143
Number of banks
2,398
1,976
2,689
The table reports the coefficients and standard errors (in parentheses). The dependent variables are as follows: the risk assets is the ratio of risky assets to total assets, the non-performing loans is the ratio of impaired loans to gross loans, the Z-score is a measure of a bank’s probability of insolvency. The explanatory variables are as follows: the slope is the long-term interest rate minus the short-term interest rate, GDP growth is the economic growth of the respective country, and Domestic credit is the amount of credit supplied by the financial institutions in the respective country. Furthermore, observations is the amount of observations included in the sample, R-squared is the amount that the explanatory variables are able to explain in the dependent variables, and the number of banks is the amount of unique banks included in the sample. ***,**, and * are the statistical significance at the 1%, 5% and 10% level, respectively.
Table 4.
Estimating equation (1) using the same sample, limited controls, and fixed effects
(1)
(2)
(3)
Variables
Risk assets
Non-performing loans Z-score
Slope
-1.221***
0.351***
0.766***
(0.074)
(0.023)
(0.030)
GDP Growth
0.251***
-0.015
0.123***
(0.035)
(0.011)
(0.014)
Domestic credit
-0.025**
0.005
0.133***
(0.011)
(0.004)
(0.005)
Constant
93.764***
1.545***
-21.885***
(1.085)
(0.341)
(0.447)
Observations
6,177
6,177
6,177
R-squared
0.198
0.138
0.579
Number of banks
1,781
1,781
1,781
The table reports the coefficients and standard errors (in parentheses). The dependent variables are as follows: the risk assets is the ratio of risky assets to total assets, the non-performing loans is the ratio of impaired loans to gross loans, the Z-score is a measure of a bank’s probability of insolvency. The explanatory variables are as follows: the slope is the long-term interest rate minus the short-term interest rate, GDP growth is the economic growth of the respective country, and Domestic credit is the amount of credit supplied by the financial institutions in the respective country. Furthermore, observations is the amount of observations included in the sample, R-squared is the amount that the explanatory variables are able to explain in the dependent variables, and the number of banks is the amount of unique banks included in the sample. ***,**, and * are the statistical significance at the 1%, 5% and 10% level, respectively.
an increase in the cost of debt, more entities are prone to default thereby increasing the non-performing loans. The last dependent variable, the Z-score, is also negative. A higher Z-score indicates that a bank incurs less risk and is more stable, but in this case I multiplied the Zscore with -1 so that a higher Z-score implies more bank risk-taking. In the sample used it can be seen that an increase in the slope of the yield curve, increases the Z-score and therefore increasing bank risk-taking. All in all, the general conclusion is that bank risk-taking increases with an increase in the slope of the yield curve. Even though there is an increase in risk free alternative investments, there are more impaired loans and banks are more prone to default with a higher slope of the yield curve.
16 Table 5.
Estimating equation (1) using the same sample, all controls, and fixed effects
(1)
(2)
(3)
Variables
Risk assets
Non-performing loans
Z-score
Slope
-0.846***
0.202***
0.718***
(0.072)
(0.022)
(0.027)
Size
-10.610***
0.196
-1.536***
(0.591)
(0.180)
(0.219)
Capitalization
0.158***
-0.065***
-0.627***
(0.053)
(0.016)
(0.020)
Profitability
1.099***
-1.101***
0.099*
(0.139)
(0.043)
(0.052)
Efficiency
0.053***
-0.023***
0.020***
(0.005)
(0.002)
(0.002)
Off-balance sheet items
3.271***
-0.434***
0.244***
(0.213)
(0.065)
(0.079)
GDP Growth
0.116***
0.007
0.069***
(0.034)
(0.010)
(0.012)
Domestic credit
0.071***
-0.009**
0.123***
(0.012)
(0.004)
(0.004)
Constant
182.572***
8.039***
1.779
(7.377)
(2.251)
(2.732)
Observations
6,090
6,090
6,090
R-squared
0.336
0.285
0.700
Number of banks
1,769
1,769
1,769
The table reports the coefficients and standard errors (in parentheses). The dependent variables are as follows: the risk assets is the ratio of risky assets to total assets, the non-performing loans is the ratio of impaired loans to gross loans, the Z-score is a measure of a bank’s probability of insolvency. The explanatory variables are as follows: the slope is the long-term interest rate minus the short-term interest rate, size is the logarithm of the total assets, capitalization is the ratio of equity to total assets, profitability is the ratio of profit before tax to total assets, efficiency is the ratio of total revenue to total expenses, balance sheet items is the logarithm of off-balance sheet items, GDP growth is the economic growth of the respective country, and Domestic credit is the amount of credit supplied by the financial institutions in the respective country. Furthermore, observations is the amount of observations included in the sample, R-squared is the amount that the explanatory variables are able to explain in the dependent variables, and the number of banks is the amount of unique banks included in the sample. ***,**, and * are the statistical significance at the 1%, 5% and 10% level, respectively.
higher probability of insolvency) with an increase of the slope in comparison to relatively poor capitalized banks. This means that relatively high capitalized banks are more risk taking than poorly capitalized banks. Therefore, the evidence to support the theory that poorly capitalized banks take more risks because they are searching for return is insufficient.
As is discussed earlier, there might be a potential for a bad control problem with the profitability. Arguably, a higher slope could cause more profitability thereby creating a bad control problem, but it could also work via the risk-taking strategy and therefore the profitability would not be a bad control. The inclusion of the control variable profitability does not alter the coefficient of the slope by much; it only increases the R-squared thereby raising the amount of movement explained. Therefore, I do not recognize profitability as a bad control and include it in the analysis.
An interesting finding is that the results suggest that the GDP growth on non-performing loans is not significant (see Table 3, 4 and 5), whereas Beck, Jakubik & Piloiu (2013) find that the real GDP growth on non-performing loans is the main driver in the past decade. This implies that even though the currency is the same, the price levels are much different.
18 Table 6.
Estimating equation (1) with the sample before and after 2008 with fixed effects
(1) (2) (3) (4) (5) (6)
Variables Risk assets Non-performing loans Z-score Risk assets Non-performing loans Z-score
Slope -0.134** 0.355*** 1.997*** -1.175*** 0.409*** -0.081*** (0.058) (0.114) (0.054) (0.082) (0.037) (0.020) Size 2.457*** -0.991** 4.253*** -9.631*** 1.704*** -1.921*** (0.209) (0.442) (0.188) (0.393) (0.229) (0.099) Capitalization 0.099*** -0.057 -0.676*** -0.143*** 0.052*** -0.573*** (0.016) (0.035) (0.014) (0.028) (0.015) (0.006) Profitability 0.023 0.040 0.030 0.616*** -0.462*** 0.082*** (0.045) (0.119) (0.036) (0.079) (0.025) (0.006) Efficiency -0.009** 0.002 -0.003 0.034*** -0.011*** 0.006*** (0.004) (0.004) (0.002) (0.003) (0.002) (0.001)
Off-balance sheet items -0.209*** -0.445** 0.273*** 1.423*** -0.570*** -0.032
(0.077) (0.180) (0.069) (0.121) (0.087) (0.031) GDP Growth -0.011 -0.267*** 0.228*** 0.019 0.012 0.229*** (0.028) (0.082) (0.026) (0.011) (0.013) (0.003) Domestic credit 0.020*** 0.004 -0.046*** 0.030*** -0.001 0.082*** (0.005) (0.009) (0.004) (0.009) (0.006) (0.002) Constant 62.146*** 21.579*** -58.241*** 202.729*** -14.245*** 17.062*** (2.647) (5.217) (2.358) (5.535) (3.121) (1.401) Observations 9,852 1,248 11,977 10,057 5,460 11,808 R-squared 0.031 0.098 0.369 0.149 0.162 0.671 Number of banks 2,049 522 2,300 2,237 1,928 2,519
survivors. Secondly, the risk taking behavior of banks decreased because banks who took too much risk got punished, thus giving less incentive to take those risks again. Therefore, the impact of a change in the slope on the Z-score is consistent with the expectations. All in all, the bank risk-taking behaviour decreased after the crisis of 2008.
Conclusion
In this paper I study the effect of the slope of the yield curve on bank risk-taking by performing a regression analysis using three different dependent variables based on 2701 unique banks over the period from 2001 to 2013. From previous literature it is known that the interest rate environment has a significant impact on bank-risk taking. I try to contribute to the existing literature by questioning the impact of the difference between the long- and short-term interest rates, namely the slope of the yield curve, on bank risk-taking.
To find the impact of the slope of the yield curve on bank risk-taking, I took a similar methodology as Delis & Kouretas (2011) by using the risk assets and non-performing loans as risk measures. However, instead of the interest rates I used the slope of the yield curve. Also, I use an additional risk measure, the Z-score, which is explained in Lepetit & Strobel (2013). The results of the regression analysis suggest that the slope of the yield curve has a negative impact on risk assets. This implies that the risk taking behaviour decreases with an increase of the slope. On the non-performing loans the results suggest that there is an increase of risk-taking by banks following an increase in the slope of the yield curve. I believe that the reason for this is that the risk free alternative has relatively become more attractive. But because of the attractiveness of the alternative, the quality of loans is dropping which results in the positive effect of the slope on the non-performing loans. The impact of the slope on the Z-score suggests that an increase in the slope is paired with an increase in the probability of default. Overall, because the Z-score indicates that there is a higher probability of default if the slope goes up, I believe the effect of non-performing loans is stronger than the effect of switching to the risk free alternative. Therefore, the slope of the yield curve has a positive effect on bank risk-taking.
From the interest rate data and Sibbertsen, Wegener & Basse (2014) I brought the attention towards a possible change in bank risk-taking behaviour as a result of the crisis. The results of this analysis pointed out that banks after 2008 are overall more risk averse than before the crisis.
20 research that I would like to make is the use of the duration gap. The duration gap would be an excellent instrument in combination with the non-performing loans to measure the effect of the slope of the yield curve. However, I do believe it is more appropriate to use it in combination with a sample consisting of banks from the United States because more information is available seeing as the sample size in this research got reduced by a lot because of the non-performing loans measure.
References
Angrist J.D. & Prischke, J.S., 2008. Mostly Harmless Econometrics: An Empiricist’s Companion. Princeton University Press, Princeton.
Beck, R., Jakubik, P., Piloiu, A., 2013. Non-performing loans: what matters in addition to the economic cycle? (Working Paper No. 1515). Retrieved from European Central Bank website: https://www.ecb.europa.eu/pub/pdf/scpwps/ecbwp1515.pdf
Borio, C., Zhu, H., 2011. Capital regulation, risk-taking and monetary policy: A missing link in the transmission mechanism? Journal of Financial Stability, 8:236-251.
Boyd, J.H.,Graham, S.L., 1986. Risk, Regulation, and Bank Holding Company Expansion into Nonbanking. Quarterly Review – Federal Reserve Bank of Minneapolis. 10(2):2-17. Buch, C.M., Eickmeier, S., Prieto, E., 2014. In search for yield? Survey-based evidence on bank
risk-taking. Journal of Economic Dynamics & Control, 43:12-30.
De Nicolò, G., Dell’Ariccia, G., Laeven, L.,Valencia, F. 2010 Monetary Policy and Bank Risk Taking. IMF Staff Position Note.
Delis, M.D., Kouretas, G.P., 2011. Interest rates and bank risk-taking. Journal of Banking & Finance, 35(4):840-855.
Dell’Ariccia, G., Laeven, L., Marquez, R., 2014. Real interest rates, leverage, and bank risk-taking. Journal of Economic Theory. 149:65-99.
Duprey, T., Lé, M., 2015. Bankscope Dataset: Getting Started. Retrieved from Social Science Research Network, http://ssrn.com/abstract=2191449.
Estrella, A., Mishkin, F.S. 1998. Predicting U.S. recessions: Financial variables as leading indicators. Review of Economics and Statistics, 80(1):45-61.
Fiordelisi, F. Marques-Ibanez, D., Molyneux, P., 2011. Efficiency and risk in European banking. Journal of Banking & Finance. 35(5):1315-1326.
Hakenes, H., Schabel, I., 2011. Bank size and risk-taking under Basel II. Journal of Banking & Finance. 35(6):1436-1449
Männasoo, K., Mayes, D.G., 2009. Explaining bank distress in Eastern European transition economies. Journal of Banking & Finance. 33(2):244-253
Ozturk, H., Pereira, L.F.V.N., 2013. Yield Curve as a Predictor of Recessions: Evidence from Panel Data. Emerging Markets Finance & Trade, 49(5): 194–212.
Rajan, R.G., 2005. Has financial development made the world riskier? (Working Paper No. 11728). Retrieved from National Bureau of Economic Research website:
http://www.nber.org/papers/w11728
