1
MSc Finance Thesis
Monetary Policy and Bank Risk Taking
Shiqi Xia (S2637375)
Supervisor: Prof. Dr. Sumru Altug
2
Monetary Policy and Banking Risk Taking
Abstract
This paper uses approximately 299 observations of Eurozone banks from 2006 to 2008 to estimate the relationship between monetary policy and bank risk taking. The results show that monetary policy does influence bank risk taking. Typically, the capital adequacy ratio plays an important role in this relationship. For banks with a high capital adequacy ratio, bank risk taking is negatively correlated with policy rates. As the capital adequacy ratio decreases, risk shifting increases. It is likely that bank risk taking is positively related to monetary policy. Besides that, this paper also illustrates that the relationship between monetary policy and macro-prudential policy is determined by not only the macroeconomic environment but also capital adequacy ratios of the whole banking system.
Key words: Monetary policy, Bank risk taking, Capital adequacy ratio, Effect of risk
shifting, Macro-prudential policy
1. Introduction
Recently, with the eruption of the global financial crisis, debate over one of the bank risk taking channels, namely, monetary policy, has become increasingly prominent.
3
Monetary policy is the set of measures used by the central bank to control the price level and to achieve the desired level of economic activity. Generally speaking, the central bank controls policy rates in order to influence the supply of money for some macroscopic targets, such as economic growth, lower unemployment, and financial stability and so on. As commonly referred to, monetary policy falls into two types: expansionary or contractionary. An expansionary monetary policy is usually adopted in a recession, during which the central bank lowers the policy rates to increase credit and encourage the expansion of the business. A contractionary monetary policy, on the other hand, refers to the opposite measures such as increasing the interest rate and enacting stricter regulations, which is applied to slow down the inflation rate and maintain the stability of asset prices.
During the period between 2007 and 2009, the global financial crisis brought a depression of economic activity. To stimulate economic activity, central banks from most countries decreased their policy rates. Specifically, Reserve Bank of Australia (RBA) decreased policy rates from 6.75% to 3.75%, and in the year of 2008, RBA reduced its policy rate by around 2%. The Federal Reserve System (Fed) lowered the interest rate from 5.25% to 0.25% during this period. After 2008, the Federal Reserve System kept its policy rate between zero and 1%. Similarly to the Fed, the Bank of Canada (BoC)
launched the analogous adjustment. The policy rate was reduced from 4.5% in 2007 to 0.25% in 2009, and the current policy rate remains 0.5%. In the European continent, the Bank of England (BoE) and the European Central Bank (ECB) also took actions to deal with the financial crisis. The BoE cut its policy rate from 5.25% in February 2007 to 0.5% in March 2009. The ECB decreased its policy rate from 4% in 2007 to 1% in 2009, and maintained the trend of decreasing rates. The policy rate of the euro area is 0.05% at present, similar to other countries that were affected by the global financial crisis. Consequently, on one hand, the low policy rates contributed to the economic recovery. On the other hand, lower policy rates created plenty of new problems. For example, the low interest rate are said to have led to excessive bank risk taking.
4
excessive bank risk taking. Many researchers pointed out that if the central bank could increase their interest rates before the accumulation of risk, the damage resulting from the financial crisis would be minimized. Many of them believe that the low policy rates presented by the central banks to mitigate the financial crisis do actually lead to further financial crisis.
Under this situation, macro-prudential policy is a hot topic. Macro-prudential policy is proposed for mitigating the financial crisis. Compared with the traditional
micro-prudential policy, macro-prudential policy focuses on systemic risk. It is used for maintaining the financial stability and economic growth. Generally speaking, the
macro-prudential policy concerns the cross sectional dimension and the time dimension. The cross sectional dimension refers to the common risk exposure that comes from the homogeneity and correlation of the financial institutions. This common risk exposure is believed to be the main cause of the bankruptcies of many financial institutions. The time dimension refers to the characteristics of the pro-cyclical period within the financial system. The combination of horizontal dimension and time dimension forms a state of top-down management which helps to supervise the whole financial system. There is no primary tool for the implementation of the macro-prudential policy; the most common instruments are caps on debt-to-income ratio and loan loss provisions, caps on leverage, liquidity coverage ratios and so on. Each of these tools emphasizes different aspects of macro-prudential policy.
In this paper, taking into account the capital adequacy ratio of banks in a macroscopic view, I discuss the relationship between monetary policy and macro-prudential policy. In particular, I focus on how these two policies react to
influence bank risk taking. Additionally, I also research whether there is a complementary relationship or a substitutable relationship between the monetary policy and the
macro-prudential policy.
Section 2 is the literature review. Section 3 gives the theoretical analysis on bank risk taking and shows the assumptions of this paper. Section 4 introduces the
5 Section 6 summarizes the results of the paper.
2. Literature review
We often discuss bank risk taking separately to these two types of monetary policy. According to Borio and Zhu (2008), if the central bank could increase the interest rate effectively before the accumulation of risks, the damage induced by financial crisis would be less. Besides that, they also put forward a channel for the transmission of monetary policy called the “risk taking channel”. The Basel requirements, which set the minimum capital standard, have influenced behavior of banks and drawn attention from investors on the capital adequacy of banks. In this paper, the authors sum up the “capital threshold effect” and the “capital framework effect”. The risk taking channel usually refers to capital threshold effect and ignores the capital framework effect. In fact, the capital framework effect influences the way in which banks perceive, manage and price the risks. Furthermore, it influences the banks’ behavior and the implementation effect of the monetary policy.
From De Nicolò et al. (2010), expansionary monetary policy leads to greater bank risk taking. As banks are always searching for yield and profits, they try to ensure the benefits of leverage maximization. The authors show a complex relationship between monetary policy and bank risk taking. They corroborate that an opposite risk-taking effect exists when financial agencies operate with limited liability. The well-capitalized banks choose to increase the risk taking when the policy rate is low. However, the
6
interest rates on risk assets is lower for banks with higher capital and is increased for banks with higher off-balance sheet items after the tests.
To find out what exactly leads banks to take more risks when monetary policy rates are low, Valencia (2014) has introduced a dynamic bank model. In this model, the policy rate can lead to profit in two ways. One is that banks may reduce the financing costs. The other one is that monopolistic banks may extract the entire increased surplus from
borrowers when rates are low because the borrowers’ opportunity costs are fixed. Because of the constraints of limited liability, banks would lever up excessively to finance new loans, resulting in excessive risk taking. In another incidence, if the interest rate is lower, the loan and the leverage will be higher, thus the limited liability constraints will be more and more important. In this paper, under better capital requirement, the author introduces some regulations such as capital requirements and loan-to-value (LTV) caps to understand their role in diminishing bank risk taking.
Numerous researchers have analyzed the relationship between monetary policy and bank risk taking from an empirical perspective. Jimenez et.al (2007) tested the data from the Spanish banking system, controlling for the macroeconomic conditions and the banks’ characteristics. They concluded that the low short-term interest rate would prompt the banks to reduce their lending standards. Although they limited the market to a single country, the results supported the franchise value paradigm.
Ioannidou et.al (2009) analyzed data from Bolivia, whose banking system is fully affected by the rate of federal funds. They found that the decrease in US federal funds rate raised the probability of default on individual bank loans. The bank increased the loans with subprime credit ratings and offered loans to riskier borrowers with current or past non-performance. Although the bank decreased the bank lending standard, the loan spreads did not increase. Finally, the low federal funds rate increased the probability of default and then increased the chances of bank risk taking. In this paper, the authors treated the US federal funds rate as an exogenous condition and they found similar results as others.
7
short-term interest rate and bank risk taking. After the test, they concluded that a low interest rate over a long period increased bank risk taking. This conclusion is similar to the results of most researchers.
In conclusion, from the theoretical horizon, most scholars have common views about the relationship between the monetary policy and bank risk taking. They believe that the easing of monetary policy induces high bank risk taking. However, there is no common view about the influence of the capital adequacy ratio. Moreover, most of the scholars focus on the relationship between money policy and bank risk taking without considering the effects of macro-prudential policy. In this paper, I investigate the relationship between monetary policy and bank risk taking firstly. On that basis, l therefore analyze the
relationship between monetary policy and macro-prudential policy.
3. Theoretical Analysis
From the former papers, it is apparent that there exists a negative relationship between the monetary policy rate and bank risk taking. Four effects contribute to the negative relationship.
1. The effect of valuation, income and cash flows of banks. Specifically, the increase in the interest rate results in the reduction of the value of capital and mortgages, revenue and profit. Furthermore, the risk perception of the bank becomes weaker when the risk tolerance is stronger. Banks cannot estimate the probability of default, loss given default, volatility and correlation of risks.
2. The effect of asset substitution. The low real yield on the safe assets leads to a downside tendency in their weight in the portfolios of banks. For risk-neutral banks, they prefer to increase the demand for risky assets until the returns on both two types of asset are equalized, which is called the equilibrium situation. For risk-averse banks, they also increase the demand for risky assets and a decrease in the proportion of low risk assets.
8
long term, financial institutions cannot achieve their promised goals, and they would lose their customers. Thus they may have to renegotiate their long-term commitments. It can therefore be seen that the focus on risky assets makes the financial institutions face more risk taking.
4. The effect of leverage. In general, financial institutions target constant or pro-cyclical leverage ratios. When financial institutions expand their balance sheets, the leverage ratios increase. Under the MM Theory, the selection of the investment project is independent of the financing method. When faced with exogenous shocks, financial institutions react by adjusting their current capital constructions instead of distributing dividends or raising new capital. Eventually, monetary policy easing will boost asset prices. Meanwhile, the demand for assets increases. Correspondingly, bank equity increases and the leverage ratios decrease. This reaction boosts asset values and so on. The bank system is much riskier when it is exposed to negative shocks.
Another effect has a great influence on bank risk raking. It is always leading to a positive relationship between monetary policy and bank risk taking. This is the effect of risk shifting. The limited liability and asymmetric information contribute to the effect of risk shifting. Under these two conditions, risk-neutral banks tend to operate like
risk-loving banks and they prefer risky investments with higher yields in case of positive outcomes to prudent investments with a higher net present value. Because of the limited liability, these banks do not have to internalize the losses, so they can impose the losses on depositors and bondholders. According to “skin-in-the-game” effect, the more the banks lose in failure, the less severe the moral hazard problem. The more of the bank’s own capital is required, the more prudently the bank will invested.
What’s more, the effects referred to in the last part interact with each other, which contributes to the final complicated relationship between monetary policy and bank risk taking.
9
GDP growth rate reached its lowest point and at the same year, loan growth rates experienced a little increase. After the financial crisis, loan growth rates were mostly below GDP growth rates. In 2012 and 2013, both the GDP growth rate and the loan growth rate were negative.
Figure 1. GDP Growth Rates and Loan Growth Rate
Figure 2 shows the relationship between the euribor rates and the non-performing loans ratio from 2006 to 2014. The non-performing loans ratios are the average values of all banks in the same year in the sample. It is obvious that there exists a negative
relationship between these two factors.
Figure 2. Euribor rates and Non-performing loans ratios
Figure 3 illustrates that the euribor and the risk weighted assets ratio is positively
-20.00% -10.00% 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 2006 2007 2008 2009 2010 2011 2012 2013 2014
GDP Growth Rates Loan Growth Rates
10
correlated from 2006 to 2014. This result is the opposite with Figure 2. The decrease of the euribor rate is not accompanied by an increase in the risk assets ratio.
Figure 3. Euribor rates and Risk weighted assets ratios
Figure 4 is the distribution of the capital adequacy ratios from 2006 to 2014. I use the total capital ratio to present capital adequacy in this paper. As shown, the distribution of CAR in this diagram is dispersed. Different capital adequacy ratios play different roles in risk shifting.
Figure 4. CAR Scatter Diagram (%)
Based on the above analysis, in this paper, I assume that monetary policy and bank risk taking has the following relationship:
11
When the capital adequacy ratio is high, the effects of risk shifting are low, and monetary policy and bank risk taking are negatively correlated. With the decreasing of the capital adequacy ratio, the effect of risk rifting becomes stronger so that the negative correlation is no longer obvious. There will be a positive relationship between monetary policy and bank risk taking.
4. Methodology 4.1 Variables
4.1.1 Dependent variable indicates bank risk taking
The most common variables to indicate bank risk taking are the Expected Default Frequency, the Z-score, the risk-weighted assets ratio and the non-performing loan ratio. In this paper, following Delis et al 2011, I choose the risk-weighted assets ratio to indicate the banks’ risk taking. Compared with the risk weighted assets ratio, the EDF is the best measurement variable for bank risk taking. However, it is hard to collect this data in practice. The Z-score is a reference to the insolvency risk. It cannot reflect status of bank risk taking precisely. For robustness, I use the NPL ratio as the measurement variable. Comparing these two variables, risk weighted assets contain more kinds of assets than non-performing loans. Risk weighted assets include all bank assets except cash, government securities and balances due from other banks. However,
non-performing loans only contain loans. Risk weighted assets can reflect market risks. Non-performing loans only reflect credit risks. In the financial system, many
macroeconomic factors would induce the default of borrowers. Therefore, the market risks are more evident. Eventually, non-performing loan ratios cannot indicate the condition of bank risk taking precisely. In this paper, the RWA ratio is more appropriate than the NPL ratio as measurement variable for bank risk taking.
4.1.2 Explanatory variable indicates monetary policy
12
4.1.3 Control Variables
Considering factors which influence both monetary policy and bank risk taking, I divide the control variables into two types: variables referring to bank characteristics and variables referring to the macroeconomic environment.
The control variables referring to the bank characteristics are bank size (SIZE), return of average assets (ROAA) and equity ratio (CAP).
Bank size is the logarithm of total assets. Generally speaking, banks with large asset size have various investment and mature risk management skills. So the whole risk of the banks is low. There may be a negative relationship between bank risk taking and bank total assets. However, because of the financial crisis, the total assets of banks fell sharply in the past years. So the positive relationship is also reasonable.
Return of average assets shows the profitability of banks. On the one hand, the greater risk taking always inspires more profit. On the other hand, more risk taking may result in more non-performing loans which reduces the profit margin. So, the relationship between the ROAA and bank risk taking is uncertain.
The equity ratio is the inverse of the leverage ratio. It means a degree of
capitalization for banks. Theoretically, the larger the equity ratio is, the less bank risk taking occurs. The steadier the bank is, the less risks the bank will take. Bank risk taking and equity ratio are negatively correlated.
The control variable referring to the macroeconomic environment is GDP growth rate. When the macroeconomic environment is weak, the market will be unstable and banks should take much more risks. Monetary policy is eased at the same time. Instead of investing in riskier assets, the main reason for banks to take more risk is that the assets held by banks become much riskier than before.
4.1.4 Interaction term
13
ratio to express capital adequacy ratio. When the total capital ratio is less than 10% in this sample, DUM_CAR equals 0. Otherwise, it equals 1.
4.2 Model construction
Following Delis et al 2011, there exists a static model:
RWAi,t=β0+β1ERi,t+β2ERi,t*DUM_CARi,t+β3SIZEi,t+β4ROAAi,t+β5CAPi,t+β6GDPi,t+
μi+εi,t (1)
μi means the individual effects and εi,t is random disturbance term. In the
experimental result, they come down to the constant term.
Delis et al 2011 have confirmed that the risks of banks are continuous. In this paper, apart from the static model, I will use the dynamic model to estimate the data. In a general way, return on average assets and equity ratios are endogenous. The dynamic model is built to solve the endogenous problem. The dynamic model is constructed as follows:
RWAi,t=β0+β1ERi,t+β2ERi,t*DUM_CARi,t+β3SIZEi,t+β4ROAAi,t+β5CAPi,t+β6GDPi,t+
+ρRWAi,t-1μi+εi,t (2)
For robustness, I will use the NPL ratio instead of the RWA ratio in both models to measure the relationship between bank risk taking and monetary policy.
4.3 Data
14
Variable Mean SD Min Max
summary statistics for the full sample
RWA 0.62558 0.18467 0.05719 1.45016 NPL 0.08967 0.06843 0.00001 0.65122 ER 0.01825 0.01727 0.00209 0.04644 CAR 0.16846 0.08860 0.00130 1.42100 SIZE 3.27211 1.05845 1.46398 6.48644 ROAA 0.00399 0.01079 -0.34028 0.09790 CAP 0.09989 0.04916 -0.00359 0.77849 GDP 0.00503 0.03090 -0.08864 0.06463
SIZE: Logarithm of Total assets (mil USD)
Table 1. Descriptive Statistics of Variables
4.4 Empirical Results
Firstly, I select the LM method to estimate these two models. Normally, pooling the data assumes that there is no heterogeneity. It is impossible to have balanced panel data. Different banks have different characteristics over time. So pooling the data is not the best choice here.
The balanced panel refers to two types of model: the fixed effects model and the random effects model. Before the data analysis, I compare these two types of model and decide which is better. Table 2 shows the comparison of the two effects model. According to Hausman test, P value is statistically significant. It means that the fixed effects model is appropriate.
Effects Stat d.f. Prob.
Cross-section random 167.801612 6 0.0000
Table 2. Effect Test of static model
Regression (1) illustrates the relationship between risk weighted assets ratio and euribor rates with other control variables. Table 3 is the result of the estimation.
15
Variable Coefficient Std. Error t-Statistic Prob.
ER 5.121 0.232 22.065 0.000 ER_DUMCAR -3.181 0.235 -13.561 0.000 SIZE -0.334 0.023 -14.500 0.000 ROAA -0.005 0.002 -2.511 0.012 CAP -0.002 0.001 -1.898 0.058 GDP -0.004 0.001 -4.777 0.000 C 1.694 0.081 20.969 0.000
R-squared 0.817 Mean dependent var 0.626
Adjusted R-squared 0.794 S.D. dependent var 0.185 S.E. of regression 0.084 Sum squared resid 16.758
F-statistic 35.114
Durbin-Watson stat 0.982 Prob(F-statistic) 0.000
Table 3. Result of static model
Table 3 shows that the euribor rate is positively related to risk weighted assets. It means that, overall, the decrease of the euribor rate does not lead to an increase of risk weighted assets in these banks. The result is the opposite to the findings of Delis.et al 2011. Because of the effects of risk shifting, this consequence is also reasonable. In this model, after taking into account the capital adequacy ratio, the interaction term is
negatively correlated with the risk weighted assets ratio. It means that for banks with high total capital ratio, the risk weighted assets ratio is negatively related to the euribor rates. To benefit from the dummy variable, I leave the banks with high capital adequacy ratio in this estimation. In the assumption, I assume that the effects of risk shifting for the banks with a high capital adequacy ratio is relatively low, and the monetary policy and bank risk taking are therefore negatively correlated. This result coincides with the assumption. From Table 3, the banks with higher capital ratio are more likely to take more risk as the euribor rate increases. When the capital adequacy ratio is low, the effects of risk shifting become stronger, and the correlation of monetary policy and bank risk taking tend to be positive. In this estimation, the coefficient of the euribor rates is positive, which supports the assumption.
16
the dynamic model and it takes the lagged risk weighted assets ratio as an instrumental variable. Similar to regression (1), firstly, I compare two type of models. Because of the statistically significant P value, in this case, I still choose the fixed effects model.
Table 4 is the result of fixed effects tests of dynamic model.
Effects Stat d.f. Prob.
Cross-section random 529.705092 7 0.0000
Table4. Effect Test of dynamic model
The result of regression (2) shows the similar consequences compared with
regression (1). The euribor rates are positively correlated to bank risk taking. Banks with high capital adequacy ratio are seriously affected by risk shifting. The monetary policy of easing will make them take more risk. Table 5 shows that lagged risk weighted assets ratio is significantly positively correlated to the risk weighted assets ratio. The coefficient value is between 0 and 1. Following Delis et al 2011, it means that the risks of the bank are continuous. Bank risk taking is exactly influenced by the former risks of banks. In the dynamic model, GDP growth rate is not statistically significant.
Variable Coefficient Std. Error t-Statistic Prob.
RWAT 0.640 0.015 41.858 0.000 ER 1.920 0.153 12.528 0.000 ER_DUMCAR -1.180 0.150 -7.889 0.000 SIZE -0.221 0.016 -13.703 0.000 ROAA -0.001 0.002 -0.605 0.546 CAP 0.003 0.001 3.348 0.001 GDP 0.000 0.000 -0.166 0.868 C 0.889 0.061 14.537 0.000
R-squared 0.949 Mean dependent var 0.851
Adjusted R-squared 0.941 S.D. dependent var 0.387 S.E. of regression 0.065 Sum squared resid 8.744
F-statistic 127.035
Durbin-Watson stat 2.004 Prob(F-statistic) 0.000
Table 5. Result of dynamic model
17
on average assets in the static model is also negatively related to bank risk taking, which means that banks in the Eurozone got numerous profit opportunities during the period of monetary policy easing. Low policy rates fuelled the profits of the banks. But in the dynamic model, it is not statistically significant. Finally, equity ratio and GDP growth rate are not statistically significant in static and dynamic model separately.
To sum up, the results of the estimation support the assumption I proposed
previously. It means that the capital adequacy ratio has a significant impact on bank risk taking, which can change the relationship between monetary policy rate and bank risk taking. This conclusion is similar with Dell’ Ariccia et al 2010 and De Nicolo et al 2010.
4.5 Robustness
4.5.1 Robustness of Non-performing loan ratio
There are many ways to measure bank risk taking, in this section, I choose the non-performing loan ratio as the dependent variable indicating bank risk taking. Similar to regression (1) and regression (2), in this section, I only change the dependent variable.
The static model here is:
NPLi,t=β0+β1ERi,t+β2ERi,t*DUM_CARi,t+β3SIZEi,t+β4ROAAi,t+β5CAPi,t+β6GDPi,t+μi
+εi,t (3)
The dynamic model here is:
NPLi,t=β0+ρNPLi,t-1+β1ERi,t+β2ERi,t*DUM_CARi,t+β3SIZEi,t+β4ROAAi,t+β5CAPi,t+β 6GDPi,t+μi+εi,t (4)
As before, I test the effects of the model. Table 6 shows that the fixed effect model is better in this regression.
Effects Stat d.f. Prob.
Cross-section random 118.568725 6 0.0000
Table 6. Effect tests of Static Model about NPL ratio
Table 7 is the result of the estimation based on non-performing ratio. Comparing Table 3 and Table 7, it is clear that the impact on the non-performing loan ratio is much more significant than it is on the risk weighted assets ratio. The euribor rates are
18
loan assets of banks more than the total risk weighted assets. The monetary policy is thus closely related to credit risks of banks. Considering the interaction term ER_DUMCAR, the P-value is larger than 0.05. The ER_DUMCAR is not significantly statistically related to the non-performing ratio. The large P-value states that the effect of risk shifting is not apparent here. Returns on average assets are negatively related to the non-performing loans ratio. This result is similar to the results from the risk weighted assets ratio in the static model. However, the bank size is positively relevant to the NPL ratio. The results show that the larger banks are more likely to have bad loans.
Variable Coefficient Std. Error t-Statistic Prob.
C -0.190168 0.038650 -4.920305 0.0000 ER -1.491626 0.111019 -13.43579 0.0000 ER_DUMCAR 0.190015 0.112188 1.693719 0.0904 GDP 0.002280 0.000364 6.257483 0.0000 SIZE 0.086493 0.011026 7.844415 0.0000 ROAA -0.019576 0.000859 -22.79044 0.0000 CAP 0.002980 0.000422 7.068632 0.0000
R-squared 0.695609 Mean dependent var 0.089667 Adjusted R-squared 0.656827 S.D. dependent var 0.068428 S.E. of regression 0.040086 Sum squared resid 3.834021 F-statistic 17.93621 Durbin-Watson stat 0.717773 Prob(F-statistic) 0.000000
Table 7. Result of Static Model about NPL ratio
To further argue the relationship between monetary policy and the non-performing loan ratio, I also estimate the dynamic model here. As seen in table 8, I still choose the fix effect model.
Effects Stat d.f. Prob.
Cross-section random 184.690748 7 0.0000
Table 8. Effect tests of Dynamic Model about NPL ratio
19
monetary policy influences loan assets more sensitively. The interaction term is not statistically significant. So the coefficient here hardly reveals the effect of risk shifting. The bank size has the same trend with the non-performing ratio. This sign of the
coefficient validates again that banks with more assets in the Eurozone are more likely to have non-performing loans. Returns on average assets and the non-performing loans ratio have a negative relationship, which is the same as the other estimations above. The equity ratio and GDP growth rate are not statistically significantly in the dynamic model based on the non-performing loans ratio. These two results are different from those in the static model. So the relationship between the non-performing loan ratio, the equity ratio, and GDP growth rate should be discussed in the specific conditions.
Variable Coefficient Std. Error t-Statistic Prob.
C -0.135527 0.027227 -4.977675 0.0000 NPLT 0.869145 0.013740 63.25484 0.0000 ER -0.376695 0.072395 -5.203353 0.0000 ER_DUMCAR 0.113918 0.072377 1.573953 0.1156 SIZE 0.049369 0.007783 6.342906 0.0000 ROAA -0.010624 0.000551 -19.28631 0.0000 CAP 0.000451 0.000279 1.612738 0.1070 GDP -0.000224 0.000228 -0.982501 0.3260
R-squared 0.899009 Mean dependent var 0.094197 Adjusted R-squared 0.884243 S.D. dependent var 0.069814 S.E. of regression 0.023753 Sum squared resid 1.176932 F-statistic 60.88308 Durbin-Watson stat 1.819150 Prob(F-statistic) 0.000000
Table 8. Result of Dynamic Model about NPL ratio
To sum up, when choosing the non-performing loan ratio as the dependent variable indicating bank risk raking, the effect of risk shifting is difficult to test. However, the robustness analysis confirms that there exists a negative relationship between bank risk taking and monetary policy. This conclusion is consistent with the assumption in Section 3.
4.5.2 Robustness of the Dummy Variable
20
According to Table 1, the average value of the total capital ratio is 16.85%. It is better to choose reference levels both larger and smaller than this number. In this paper, I choose the reference levels at 10%, 15%, 20% and 25%. The results of these regressions are similar to former results. The interaction terms are negatively related to the risk weighted assets ratio. It means that the assumption is right. Banks with relatively high capital adequacy ratios are more likely to take more risks than others in the face of easing monetary policy.
Figure 5.Effect of risk shifting
Figure 5 shows that under the reference levels of 10%, 15% 20% and 25%, for banks with a high capital adequacy ratio, the euribor rates are all negatively correlated with the risk weighted assets ratio. However, as can be seen in the figure, the trend of change is not linear. As the decreasing in capital adequacy ratio, the effect of risk shifting is increasing. Besides that, the positive relationship between the euribor rates and the risk weighted assets ratio is more and more significant. This conclusion supports the result of the regression in the data analysis.
4.5.3 Robustness of Dynamic Model
The hidden assumption of the dynamic model in this paper is that bank risk raking is persistent. To support the continuity of risk taking, in this part, I will estimate the model by GMM/DPD method. Following regression (2),the control variables are the equity ratio, the return on average assets, the GDP growth rate, and total bank assets. I treat the equity
-4 -2 0 2 4 6 8 10 10% 15% 20% 25%
21
ratio and the return on average assets as endogenous variables. The set of instruments is RWAi,t-1, CAPi,t, ROAAi,t. Bank size is treated as predetermined variable. The bank size
determines the level of bank risk taking. This implies that the bank size is also a valid instrument. Table 9 shows the result of the estimation.
The coefficients of the euribor rates and the interaction term have the same signs as those in Table 5. They are also statistically significant. These two variables support viewpoints sharing the assumption. Bank size is negatively correlated with the risk weighted assets ratio. Greater bank assets can mitigate bank risks. Under the GMM/DPD method, the GDP growth rate, the return on average assets, and the equity ratio do not significantly affect bank risk taking. However, the coefficient of the lagged risk weighted assets ratio is over 1. This result is abnormal. The P-value of the Sargan test is less than 0.05, which indicates that the model is over-identified. The instrument variables are not appropriate in this model. With the GMM/DPD method, the estimation based on non-performing loan ratios has similar outcomes.
Variable Coefficient Std. Error t-Statistic Prob.
RWA(-1) 1.061206 0.054347 19.52640 0.0000 ER 3.620124 1.263182 2.865878 0.0042 ER_DUMCAR -4.450606 1.467595 -3.032585 0.0025 SIZE -0.280254 0.060306 -4.647223 0.0000 ROAA 0.003642 0.002427 1.500798 0.1336 CAP 0.000492 0.002185 0.224944 0.8220 GDP 0.001795 0.000725 2.475048 0.0134
Mean dependent var -0.025012 S.D. dependent var 0.073795 S.E. of regression 0.110940 Sum squared resid 25.67398
J-statistic 62.54575 Instrument rank 31
Table 9. Result of GMM/DPD Method
Considering this unusual consequence, I attribute the causes to the dataset in 3 aspects.
22
uses 15000 observations to do the estimation. From that paper, the distribution of banks are relatively uniform. Their conclusion is more rational. The huge difference from the dataset contributes to the large coefficient and small P value of the Sargan test.
Secondly, in this paper, I choose banks from different countries in the euro area. They apply the same regulations formulated by the European Central Bank, overall. However, individually, each country draws up its own regulations to supervise banks. Banks should accept regulations not only from European Central Bank but also from their own central banks. Regulations from home countries strongly influence the operation and management of banks. The endogeneity thus cannot be measured easily. To get more precise conclusions, it is better to collect data by country. Altunbas et al 2010 and
Jimenez et al 2008 both collected data from the same country. After ensuring the identical external factors, the endogeneity is easily can be measured. In this paper, I discuss the relationship between monetary policy and bank risk taking in the euro area, so I cannot collect data from only one country. The endogeneity cannot be estimated accurately, but the coefficients of the euribor rates and the interaction term can verify the assumption sufficiently.
Finally, another reason that can be taken to explain this phenomenon is that the time period of the dataset impacts the estimation. The time period in this paper lasts from 2006 to 2014. It contains the period of the latest financial crisis. During the financial crisis, central banks decreased policy rates sharply. The euribor 3-month rate stays around 0.2% nowadays. Banks suffered great losses during this period. In this particular situation, the relationship between monetary policy and bank risk taking is greatly affected by the financial crisis. This situation can also induce the unusual results of the estimation.
5. Monetary policy and Macro-prudential policy
23
ratio of banks, the impact from monetary policy is different. The positive effects and negative effects are coexisting.
As the contagion of the financial crisis, macro-prudential policies have been used by vast numbers of financial and supervision organizations. For Galati and Moeessner (2011), the main purpose of macro-prudential policy is to maintain stability of the whole financial system. Supervisory organizations take many prudential tools to limit the systemic risk and prevent economic entities from being damaged. The capital requirement is one of the common tools used in the macro-prudential policy.
For central banks, on the one hand, they formulate the monetary policy. On the other hand, they should be supervised by other institutions via macro-prudential policy. As can be seen in Figure 6, they focus on different areas with the same purpose, which is
maintaining the financial systemic stability. On the whole, the complementary
relationship between monetary policy and macro-prudential policy has been approved by more and more researchers.
Figure 6 introduces the relationship between monetary policy and macro-prudential policy. In this paper, I take the capital adequacy ratio as a tool of the macro-prudential policy, and discuss their coordinated relations specifically in this section.
24
of the whole bank system to total assets. Ra delegates the maximum value of the systemic
risk weighted assets ratio. The assumption is that the risk weighted assets ratio is larger than 0. πhigh and πlow are the upper limit and floor limit of the inflation ratios, which the
financial system can accept. The lineπ=πlow, π=πhigh, RWA Ratio=Ra and horizontal axis
divide the coordinate system into 6 parts.
Comparing the macro-prudential and traditional micro-prudential regulation, although they focus on different scopes, they both contribute to financial stability. They both use capital regulation, loan loss reserves, loan-to-value ratios, liquidity risk indexes and stress testing as their prudential tools. In this paper, the macro-prudential policy tool is the capital regulation in the countercyclical period. The monetary policy is the euribor rate. Assuming that in the initial state, both monetary policy and macro-prudential policy are at a moderate value.
1. For part E of Figure 6, the inflation rate and the risk weighted assets rate are in the initial state. They are under a controllable range for central banks. In this situation, it is all right for central banks to keep a moderate level of monetary policy and macro-prudential policy.
2. For part A of Figure 6, the inflation rate is in a tolerance range. However, the risk weighted assets ratio exceeds the upper limit. Central banks have lost control of the whole of bank systemic risk taking. The phenomenon of financial imbalance appears. Under this condition, it is better for central banks to take stricter macro-prudential policies, such as improving the capital adequacy ratio of the whole bank system.
3. For part D of Figure 6, the inflation rate is either larger than the upper boundary or less than the lower boundary. Both conditions are out of the tolerance levels for central banks. This is a normal situation in economics. For the right part of D, the central banks could take tightened monetary policy. In the Eurozone, the ECB may increase the euribor rates. For the left part of D, the loose monetary policy is the best way to deal with the situation. Central banks should decrease the policy rates.
25
adequacy ratio. So it is better to discuss the specific relation under different capital adequacy ratios.
Firstly, when the capital adequacy ratio is lower than the reference level, bank risk taking would have a positive relationship with monetary policy. The tightened monetary policy will lead to the increase of bank risk taking. So, at this time, simply adjusting the monetary policy is not the best way to deal with problems concerning financial
imbalances and price imbalances. The restrictive monetary policy, such as increasing the euribor rates, is used to cope with inflation and the strict macroprudential policy is used to mitigate bank systemic risk taking. Since the contractionary monetary policy induces greater bank risk taking, the macroprudential policy should be much stricter. In this case, the macroprudential policy requests banks that have relative low capital adequacy ratios should increase their capital adequacy ratios more, as should the banks who play very important roles in whole bank system. In a sense, this method is as same as the analysis proposed by mainstream scholars.
Secondly, when the capital adequacy ratio of total bank system is above the reference point, bank risk taking and the monetary policy are negatively related. Under this premise, the tightened monetary policy will decrease the inflation and bank risk taking. In this case, either contractionary monetary policy or macro-prudential policy may realize price stability and financial stability at the same time. In this situation, the
monetary policy and the macro-prudential policy are substituted for each other in a way. 5. For part C of Figure 6, the inflation rate is lower than the minimum value and the risk weighted assets ratio is larger than the maximum value. This situation is also out of the control of the central bank. Opposite to part B, when the capital adequacy ratio of the whole bank system is less than a specific reference level, the easing monetary policy has a substitute relationship with the macro-prudential policy. Either the loose monetary policy or the macro-prudential policy can maintain price stability and financial stability.
26
the strict macro-prudential policy is to improve the capital adequacy ratio of the whole banking system.
6. Summary
In this paper, considering the capital adequacy ratio, I collect the data of 299 banks in the Eurozone from 2006 to 2014 and conduct an analysis of the relationship between bank risk taking and monetary policy. Based on the conclusion, I further research the relationship between monetary policy and macro-prudential policy. Either the risk weighted assets ratio or the non-performing loans ratio is selected as the measureable variable. The experimental results from both the static model and the dynamic model can support the assumption.
The experimental results conclude that:
1. The monetary policy has a significant impact on bank risk taking. The influence is related to the capital adequacy ratios of banks. When the banks have a high capital
adequacy ratio, the monetary policy is negatively related to bank risk taking. As the capital adequacy ratio decreases, the effect of risk shifting increases. To sum up, the way which the monetary policy influences bank risk taking dependent on the combined action from both the degree of the capital adequacy ratio and the interactions of the effects I mentioned in section 3.
2. The effect of risk shifting is decided by the capital adequacy ratio. The effect of risk shifting can determine the total relationship between monetary policy and bank risk taking. The negative relationship between the bank risk taking and monetary policy is rarely relevant to the capital adequacy ratio. Banks with a high capital adequacy ratio are less affected by risk shifting. The impact on the effect of risk shifting is not linear.
3. The coordinated relations between the monetary policy and the macro-prudential policy refer to not only to the economic position but also to the capital adequacy ratio. Particularly, when the inflation rate and risk weighted assets ratio are under the control level, the moderated monetary policy and macro-prudential policy are feasible. When inflation is modest and bank risk taking is exceeding the standard levels, strict
27
inflation is out of limits, monetary policy is enough. When the inflation and bank risk taking are both above the upper limits, the policy tools are dependent on the degree of the capital adequacy ratio. If the capital adequacy ratio is less than a reference point, the tightened monetary policy and macro-prudential policy should both be taken. If the capital adequacy ratio is larger than the reference point, only the tightened monetary policy is needed. When the inflation is lower than the minimum value and the bank risk taking is over the maximum value, the choice of regulator tools should also depend on the capital adequacy ratio. If the capital adequacy ratio is greater than the reference level, the easing monetary policy and macro-prudential policy should both be put forward to maintaining financial stability and price stability. Otherwise, only the loose monetary policy is enough.
28
References
1. Delis M, Kouretas G. Interest rates and bank risk-taking. Journal of Banking &
Finance. 2011; 35(4):840-855. Doi:10.1016/j.jbankfin.2010.09.032.
2. Valencia F. Monetary policy, bank leverage, and financial stability. Journal of
Economic Dynamics and Control. 2014; 47:20-38. Doi:10.1016/j.jedc.2014.07.010.
3. Borio C, Zhu H. Capital regulation, risk-taking and monetary policy: A missing link in the transmission mechanism? Journal of Financial Stability. 2012; 8(4):236-251. Doi:10.1016/j.jfs.2011.12.003.
4. Dell'Ariccia G, Marquez R, Laeven L. Monetary Policy, Leverage, and Bank Risk-Taking. IMF Working Papers. 2010; 10(276):1. Doi:10.5089/9781455210831.001.
5. De Nicolo G, Dell'Ariccia G, Laeven L, Valencia F. Monetary Policy and Bank Risk Taking. SSRN Electronic Journal. Doi:10.2139/ssrn.1654582.
6. Galati G, Moessner R. Macroprudential Policy - A Literature Review. Journal of
Economic Surveys. 2012:no-no. Doi:10.1111/j.1467-6419.2012.00729.x.
7. Jimenez G, Lopez J, Saurina Salas J. How Does Competition Impact Bank Risk-Taking? SSRN Electronic Journal. Doi:10.2139/ssrn.1582331.
8. Jimenez G, Ongena S, Peydro J, Saurina Salas J. Hazardous Times for Monetary Policy: What Do Twenty-Three Million Bank Loans Say about the Effects of Monetary Policy on Credit Risk-Taking?. SSRN Electronic Journal. Doi:10.2139/ssrn.1333538.
9. Laeven L, Levine R. Bank Governance, Regulation, and Risk Taking. SSRN
Electronic Journal. Doi:10.2139/ssrn.1142967.
10. Adrian T, Shin H. Money, Liquidity, and Monetary Policy. American Economic
Review. 2009; 99(2):600-605. Doi:10.1257/aer.99.2.600.
29
12. Peydro J, Maddaloni A. Bank Risk-Taking, Securitization, Supervision, and Low Interest Rates: Evidence from Lending Standards. SSRN Electronic Journal.
Doi:10.2139/ssrn.1571697.
13. Bond S. Dynamic panel data models: a guide to micro data methods and practice.
Portuguese Economic Journal. 2002; 1(2):141-162. Doi: 10.1007/s10258-002-0009-9.
14. De Nicolo G, Jalal A, Boyd J. Bank Risk-Taking and Competition Revisited: New Theory and New Evidence. IMF Working Papers. 2006; 06(297):1.
Doi:10.5089/9781451865578.001.
15. Ioannidou V, Ongena S, Peydro J. Monetary Policy, Risk-Taking, and Pricing: Evidence from a Quasi-Natural Experiment. Review of Finance. 2014.
Doi:10.1093/rof/rfu035.
16. Agur I, Demertzis M. Monetary Policy and Excessive Bank Risk Taking. SSRN
Electronic Journal. Doi:10.2139/ssrn.1950167.
17. Rajan R. Has Finance Made the World Riskier? European Financial
Management. 2006; 12(4):499-533. Doi:10.1111/j.1468-036x.2006.00330.x.
18. Zicchino L. A Modelof Bank Capital, Lendingandthe Macro Economy: Basel I versus Basel II*. Manchester School. 2006; 74(s1):50-77.
Doi:10.1111/j.1467-9957.2006.00517.x.
19. Montes G, Monteiro G. Monetary policy, prudential regulation and investment.
Journal of Economic Studies. 2014; 41(6):881-906. Doi: 10.1108/jes-12-2012-0173.
20. De Santis R, Surico P. Bank lending and monetary transmission in the euro area.
Economic Policy. 2013; 28(75):423-457. Doi:10.1111/1468-0327.12013.
21. Ercolani M. Essays in Panel Data Econometrics. Economic Journal. 2004; 114(496):F340-F341. Doi:10.1111/j.1468-0297.2004.00226_5.x.
30
error-components models. Journal of Econometrics. 1995; 68(1):29-51. Doi: 10.1016/0304-4076(94)01642-d.
23. Angeloni I, Faia E, Lo Duca M. Monetary Policy and Risk Taking. SSRN
Electronic Journal. Doi:10.2139/ssrn.2230335.
24. Kiss G, Vadas G. The Role of the Housing Market in Monetary Transmission - Evidence from Hungary. REUJ. 2007; 7(3):299-317. Doi: 10.1080/14616710701477938.
25. Gambacorta L. Inside the bank lending channel. European Economic Review. 2005; 49(7):1737-1759. Doi:10.1016/j.euroecorev.2004.05.004.
26. Blundell R, Bond S. Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics. 1998; 87(1):115-143. Doi:
10.1016/s0304-4076(98)00009-8.
27. Bun M, Windmeijer F. The weak instrument problem of the system GMM estimator in dynamic panel data models. Econometrics Journal. 2010; 13(1):95-126. Doi:10.1111/j.1368-423x.2009.00299.x.
28. Borio C, Tsatsaronis K. Accounting and prudential regulation: from uncomfortable bedfellows to perfect partners? Journal of Financial Stability. 2004; 1(1):111-135. Doi:10.1016/j.jfs.2004.06.001.
29. Borio C. Towards a Macroprudential Framework for Financial Supervision and Regulation? CESifo Economic Studies. 2003; 49(2):181-215.
Doi:10.1093/cesifo/49.2.181.
30. Blum J. Do capital adequacy requirements reduce risks in banking? Journal of
Banking & Finance. 1999; 23(5):755-771. Doi: 10.1016/s0378-4266(98)00113-7.
31. DINC I. Politicians and banks: Political influences on government-owned banks in emerging markets. Journal of Financial Economics. 2005; 77(2):453-479.
31
32. Gropp R, Heider F. The Determinants of Bank Capital Structure. Review of
Finance. 2010; 14(4):587-622. Doi:10.1093/rof/rfp030.
33. Xiao Y. French Banks Amid the Global Financial Crisis. IMF Working Papers. 2009; 09(201):1. Doi:10.5089/9781451873481.001.
34. Borio C. Monetary and Prudential Policies at a Crossroads? New Challenges in the New Century. SSRN Electronic Journal. Doi:10.2139/ssrn.948135.
35. Brunnermeier M, Sannikov Y. A Macroeconomic Model with a Financial Sector.
American Economic Review. 2014; 104(2):379-421. Doi:10.1257/aer.104.2.379.
36. Calem P, Rob R. The Impact of Capital-Based Regulation on Bank Risk-Taking.
Journal of Financial Intermediation. 1999; 8(4):317-352. Doi:10.1006/jfin.1999.0276.
37. Pasiouras F, Gaganis C, Zopounidis C. The impact of bank regulations, supervision, market structure, and bank characteristics on individual bank ratings: A cross-country analysis. Rev Quant Finan Acc. 2006; 27(4):403-438. Doi:
10.1007/s11156-006-0045-0.
