University of Amsterdam
Bachelor thesisCapital requirements and risk management: the effect of a financial supervisor on bank behavior
Supervisor Student
Egle Jakucionyte Terence Speijer
10465529
e.jakucionyte@uva.nl terence.speijer@student.uva.nl
Contents 1 Introduction 3 2 Literature review 6 3 Methodology 14 4 Data Analysis 17 5 Results 20 6 Discussion 23 7 Conclusion 23 Bibliography 25 Appendix 26 2
Abstract
The European Central Bank (ECB) has recently tested European Union banks subject to performance. The results indicate that several banks do not hold a sufficient level of capital and are subject to risk taking. This thesis focuses on the effects of the minimum capital requirement. In particular, whether banks in the European Union adjust risk taking, when those approach the minimum regulatory capital level. Instrumental variable regression is used. The regression results indicate that regulatory pressure does not significantly affect a bank’s risk-taking.
Introduction
Recently, the European Central Bank (ECB) has published a report, called “ECB: Aggregate Report on the Comprehensive Assessment”, in which banks were subject to performance called the comprehensive test. Several banks have failed this test and need to adjust their risk-taking. Throughout this thesis, risk-taking refers to the extent to which a bank is exposed to risk. The extent can be measured by risk-weighted assets (RWA). RWA are assets risk-weighted according to their level of
riskiness where riskier assets are assigned a higher weight. The fact that some banks have failed the comprehensive test confirmed that some banks tend to hold assets that are considered to be too risky and some banks take too many risks to resist adverse shocks to the banking system. This thesis focuses on how a regulator can prevent banks from taking too many risks. In particular, I focus on whether some banks in the European Union adjust their risk taking, when those banks approach the minimum regulatory level of capital.
The Basel Committee on Banking Supervision has developed an international framework, called Basel II. An aspect of Basel II is how to set the minimum capital requirement, which is also referred to as the capital adequacy ratio. The minimum capital requirement is defined as: the ratio of total capital to Risk-Weighted Assets (RWA). The capital adequacy ratio is set at 8% under Basel II. Accordingly, the European Parliament has agreed that banks in the European Union have to satisfy a minimum capital requirement of 8% of total capital to RWA.
A regulator, by means of a supervisor, might be needed to control the extent of a bank’s tendency to take risk. According to Demsetz et al. (1996), the role of
banking supervisors is to ensure the safety and soundness of the banking system.
Therefore, the goal is to reduce excessive risk taking in banking. A supervisor can use several tools to achieve this goal but may also need ways to enforce compliancy. This thesis discusses minimum capital requirement as a tool and regulatory pressure as a way to enforce bank’s compliancy.
Imposing a minimum capital requirement is assumed to have a desired outcome of reduced excessive risk-taking. Berger et al. (1995) and Rime (2001) note that a minimum capital requirement is motivated by the concern that a bank may hold less capital than it is socially optimal. The reason is that negative externalities
resulting from bank default are not taken into account in the level of capital that a bank chooses to hold. Taking the desired effect and motivation of minimum capital requirements into account, the probability of default decreases if a minimum capital requirement is imposed. It can be argued that banks that both reduce excessive risk taking and hold more capital have a lower probability of default. The literature review discusses how imposing a minimum capital requirement combats risk-taking in a deeper context. In short, a minimum capital requirement is considered an effective tool when imposed by a supervisor in order to help reduce a bank`s tendency to take risk.
If banks approach the minimum capital requirement, then they are subject to regulatory pressure. In other words, regulatory pressure becomes effective when banks are close to the minimum capital requirement. This thesis defines regulatory pressure as banks who deviate either by 0.2% or 0.1% from the minimum capital requirement. So the ranges are 0.4% and 0.2%, respectively. Note that this range is arbitrary and is used to limit the scope of classifying banks under regulatory pressure. The greater the range, the less effective regulatory pressure becomes. It should be noted that if a bank does not satisfy the minimum capital requirement, it can incur costs such as a supervisor imposing a fine. These costs and the effects it may have on a bank’s behavior are not within the scope of this thesis.
There are two ways to respond to satisfying the minimum capital requirement. First, a bank can increase its capital so that its capital adequacy ratio increases. Second, a bank can choose to decrease risk taking by holding relatively lower RWA. This results in a higher capital adequacy ratio. Consequently, it is more likely that a bank can satisfy the minimum capital requirement. This thesis focuses on decreasing risk-taking by using an instrumental variable regression where risk-taking is
considered to be the dependent variable. Regulatory pressure is one of the
independent variables and is tested for significance.
Earlier it was mentioned that the goal of a supervisor is to reduce excessive risk taking in banking. If a bank takes excessive risk, then it is possible that it holds an insufficient level of capital. Consequently, this could increase the probability of default. This paragraph sheds light on the relationship between bank default and risk taking. In a paper by Hellmann et al. (2000), the frequency of banking crises has increased significantly during the 1980’s and 1990’s. They discuss the importance of the effects of such a crisis. According to their theory, as a bank is more exposed to risk, the probability of increasing financial distress costs increases. To describe financial distress, Berger et al. (1995) state that a bank experiences financial distress if it has difficulty honoring its commitments (p.396). The costs of financial distress are reflected by any cost due to bankruptcy, potential loss in value, and agency problems. When many banks experience financial distress for a consecutive number of periods, it is more likely that a banking crisis emerges. Beck et al. (2006) state that a systematic banking crisis occurs when the national banking system has suffered sufficient losses (p.1584). When many banks face an increase in the probability of bank default, it is implied that the likelihood of a banking crisis increases. So, the more banks have an increased probability of bank default, the more likely a banking crisis occurs. To conclude, taking the links among financial distress costs, risk-taking and the likelihood of bank default together, the relationship between bank default and risk-taking is assumed to be positive.
Reflecting on the results of the comprehensive test, the extent of risk taking by banks, and the desired effect of imposing a minimum capital requirement, it may be clear that the aftermath of the European Union banking crisis has concerned the ECB. In particular, the inability to hold a sufficient level of capital raised questions about the extent to which a bank is exposed to risk. Therefore, the ECB introduced the comprehensive test as an indication of financial health over 130 banks in the euro area. The ECB report also explains how the test has been assessed. It consists of two parts, namely the Asset Quality Review (AQR) combined with a macro stress test. The test sketches two scenarios, a baseline and an adverse one. Under the baseline, banks experience normal economic circumstances while under the adverse scenario macroeconomic developments deteriorate quickly which can be translated to banks experiencing a simulated severe crisis. The methodology involved in the baseline scenario has been developed by the European Commission while the adverse one has
been developed by the ECB, the European Banking Authority and the European Systemic Risk Board. Two key findings are that there is €25 billion capital shortfall across 25 banks and that there is €263 billion capital depletion over a three year horizon under the adverse scenario. As mentioned earlier, the results indicate that some banks need to adjust their risk taking. Remember that imposing a minimum capital requirement can be used as a tool by a supervisor to reduce excessive risk taking. However, approaching the minimum capital requirement can be questioned. The comprehensive test does not empirically assess the effect of approaching the minimum capital requirement. Rather, it tests whether banks hold a sufficient level of capital under different scenarios. The approach used is a simulation.
Given the recent financial crisis and the results of the comprehensive test, a new discussion can be opened about the extent to which European Union banks take excessive risk. Specifically, one could question the effectiveness of approaching the minimum capital requirement. Therefore, it leads to the following research question: To what extent does approaching the minimum capital requirement incentivize banks to reduce risk-taking?
To the knowledge of this thesis, the effectiveness of approaching the minimum capital requirement imposed on European Union banks has not been examined to a greater extent. Therefore, an important contribution of this thesis is to use a panel data regression across European Union banks who approach the minimum capital requirement. In other words, this thesis seeks to verify whether European Union banks close to the minimum capital requirement reduce risk-taking due to regulatory pressure. Specifically, the regression of this thesis will focus on the use of capital requirements. Using this approach, empirical evidence can be found and the effectiveness of approaching the capital adequacy ratio can be assessed. For example, if it is found that the minimum capital requirement is insignificant in the panel data regression, then it might be implied that the current minimum capital requirement does not have its desired effect.
Literature review
So far it has been discussed that imposing a minimum capital requirement is one of the tools of a supervisor to decrease excessive risk-taking. Also, it has been discussed that regulatory pressure is a way to enforce bank’s compliancy. But no light has yet been shed upon the underlying reasons banks take risk. Therefore, the
literature review discusses three factors of risk-taking. Specifically, these are moral hazard, franchise value and competition in the banking industry.
The first factor to be discussed is moral hazard and can be defined as
following: a lender runs the risk that a borrower will engage in undesirable activities from the lender’s point a view, which makes it less likely that a loan will be repaid. The second factor to be discussed is franchise value and Hellmann et al. (2000) define the franchise value as the discounted stream of future profits (p. 149). According to Demsetz et al. (1996), a bank that holds more capital and reduces risk-taking, has a higher franchise value. Therefore, a high franchise value implies that a bank can potentially lose more profits if it takes more risk. The third factor to be discussed is competition in the banking industry. The reason why these three are classified as factors is because they seem to have a significant influence on the extent to which a bank decides to take risk as is discussed below. Notice, however, that these are not all factors determining a bank’s risk-taking. Rather, it seems that most of the banking literature regard these factors as most important. Taking these factors into
consideration can help in answering the research question. Second, different arguments for capital requirements are discussed. It was mentioned earlier that imposing a minimum capital requirement is one of the tools to reduce excessive risk taking. Thus, examining the effectiveness of approaching the minimum capital requirement throughout existing literature can help to answer the research question. A relationship between moral hazard and risk-taking can be found. By
definition, moral hazard is a form of asymmetric information. Therefore, markets are imperfect when banks are involved in moral hazard. Keeley (1990) recognized this relationship by stating that a fixed-rate deposit insurance system, which translates into moral hazard, incentivizes a bank to increase its risk-taking. Rochet (1992) explains the moral hazard argument as depositors having no incentive to control a bank`s behavior if they are insured against failures. In this argument, the depositor can be seen as the lender and the bank as the borrower, which could engage in undesirable activities from the depositor’s point of view.
In addition to the interaction between moral hazard and risk-taking, Hellmann et al. (2000) question whether deregulation in the banking industry could explain a bank`s risk-taking. According to their theory, deregulation increases competition, which results in a lower franchise value among banks. The next paragraph describes the negative relationship between franchise value and risk-taking. Also, the theory of
Hellmann et al. (2000) states that a certain, sufficient franchise value incentivizes a bank to invest prudently. If a bank has a franchise value below this certain, sufficient value, then it is more likely that a bank increases its risk-taking because it is less likely that it will invest prudently. To sum up, deregulation lowers the franchise value. This lower franchise value can reinforce the likelihood that a bank increases its risk-taking. Furthermore, Hellmann et al. (2000) state that moral hazard incentivizes a bank to increase its taking. Thus, the relationship between moral hazard and risk-taking is positive while deregulation of the banking system can lead to a higher probability of risk-taking because moral hazard can be present and has a reinforcing effect on risk-taking.
The second factor to be discussed is franchise value. Demsetz et al. (1996) elaborate on the interaction between franchise value and risk-taking. They argue that a bank with high franchise value has more to lose if a risky policy leads to insolvency as compared to a bank with low franchise value. The intuition behind this is that higher franchise values imply greater potential gains, which cannot be exploited in case of insolvency. An important result is that banks with a high franchise value hold more capital and are less exposed to risk. In other words, the higher the franchise value the more it has lose in terms of the discounted stream of future profits.
Therefore, a bank will be less incentivized to increase risk-taking when it has a higher franchise value. In addition, a high franchise value reduces moral hazard problems and banks would choose to invest prudently. Thus, there is a negative relationship between franchise value and risk taking. Therefore, a high franchise value can facilitate the role of a supervisor.
It is important to recognize that, besides moral hazard, competition among banks determines the riskiness of a bank. When many banks take too many risks, the likelihood of a banking crisis increases. Keeley (1990) argues that increased
competition leads to increased risk-taking by banks. He states that franchise values declined when competition increased. These lower franchise values can be reflected by banks holding a lower level of capital. To sum up, because an increase in the number of competitors decreases the franchise values of banks, it increases the extent to which a bank a takes risk. Therefore, the relationship between competition and risk-taking is positive.
To link the relationship between competition and risk-taking to the likelihood of a banking crisis, Beck et al. (2006) describe the concentration-stability view. This
view dictates that as a market becomes less concentrated, the probability of a banking crisis increases. To see why, the probability of a banking crisis depends partly on the ability of a supervisor to regulate banks. Arguably, it is easier to monitor a few banks in a concentrated industry than it is to monitor many banks in a less concentrated industry. Earlier, it was argued that competition increases risk-taking. Taking more risks increases the probability of bank default, and thus there is a greater likelihood of a banking crisis. If the regulation-ability of a supervisor and the number of
competitors are taken into account, it leads to the concentration-stability view. However, Boyd and De Nicoló (2005) find an opposing result and argue that less competition leads to an increase in risk. The theoretical model of Boyd and De Nicoló (2005) allows for the existence of a loan market and requires that there be the same number of banks competing for both deposits and loans. Beck et al. (2006) call this theory the concentration-fragility view. Two assumptions are needed in this view. First, the banking industry has few banks. Second, regulators are more concerned about the stability of these few banks than they would be in an industry with many banks. This leads to the notion too-big-to-fail. Under this notion, it is argued that banks tend to take more risk because they are considered to be too-big-to-fail. This is also a form of moral hazard. For example, if a bank takes too many risks and faces a high probability of default, it might receive a bailout from the regulator. In this case, the regulator is the lender and the bank is the borrower. In this scenario, the higher the probability of bank default for a bank that is considered too-big-to-fail, the higher the likelihood of a banking crisis. So, moral hazard plays a key role in the relationship between competition and risk-taking. In short, the concentration-fragility view can be summarized as following: a more concentrated banking industry increases the
likelihood of a banking crisis. Beck et al. (2006) define fragility as the probability of a banking crisis emerging.
While Keeley (1990), and Boyd and De Nicoló (2005) argue theoretically why competition interacts with riskiness, Beck et al. (2006) try to empirically assess this. In order to do so, they use concentration as a measure of competitiveness. Their main results are consistent with the concentration-stability view. They argue that it is easier to monitor an industry with few banks. Consequently, when there are many banks active in the industry there is a greater likelihood of a banking crisis occurring. However, Beck et al. (2006) are cautious about their results because of a timing problem. Also, they suggest that other measures, besides bank concentration, need to
be taken into account to assess bank competitiveness.
So far, the impact of moral hazard, franchise values, and competition in the banking industry with respect to risk-taking has been discussed. Most of the existing literature theoretically argue the relationships but empirical attempts have also been made. To summarize, risk-taking is assumed to hold a positive relationship with moral hazard, a negative relationship with franchise value, and different views exist on the relationship with competition.
As was mentioned earlier, imposing a capital requirement can serve as a tool to combat risk-taking undertaken by banks. Rochet (1992) finds support for imposing capital requirements. First, Rochet (1992) examines the impact of capital
requirements on portfolio choices of commercial banks. He elaborates on economic theory developed by Modigliani-Miller by using a portfolio model to find the different decisions faced by banks when they are maximizing their market value. In short, the Modigliani-Miller theory states that the capital structure of a firm cannot affect its market value under the assumption of a complete world with full
information. Two approaches are used. In the first approach, it is assumed that markets are complete, banks try to maximize their value and different scenarios take place such as the introduction of a deposit-rate insurance scheme and banks not being constrained by capital requirements. This approach shows that any capital
requirements are inefficient for controlling risks taken by banks. To see why, banks have to assign weights according to their level of riskiness as was mentioned earlier. This principle was referred to as RWA. One of Rochet’s (1992) results is that imposing capital requirement in a complete market’s set up does not necessarily decrease a bank’s risk-taking. In other words, the value of RWA is not minimized. Moreover, he shows that banks choose risky assets such that their exposure to risk increases as faced by their portfolio.
In the second approach of Rochet (1992), banks maximize utility and markets are incomplete. Again, different scenarios of whether capital requirements are
imposed or not are analyzed. Rochet’s (1992) result is that imposing capital
requirements can be effective if the weights assigned RWA are proportional to their systematic risk. Earlier, it was mentioned that the capital adequacy ratio was defined as the ratio of total capital to RWA. Therefore, he suggests to use so called market-based risk weights for bank loans when weights need to be assigned in RWA in order to correctly account for exposure to risk. Consequently, it is possible that risk-taking
is reduced when the suggestion is adopted. It should be noted that the methodology, developed by the Basel Committee, for computing the capital adequacy ratio has changed over time. Moreover, Rochet (1992) takes limited liability of banks into account. He argues that if a bank does not hold a sufficient level of capital, it may increase its level of risk-taking. In order to prevent increased risk-taking, Rochet (1992) suggests that a minimum capital requirement may be necessary.
In short, the first approach of complete markets showed that capital
requirements are ineffective. Therefore, imposing capital requirements do not have the desired outcome of reducing risk-taking under this approach. Moreover, the
assumption that markets are complete becomes invalid if it is shown that moral hazard is present. The second approach showed that capital requirements can be effective if banks assign market-based weights to bank loans in their RWA. In addition, if limited liability is taken into account, a capital requirement might be necessary in order to prevent risk-loving behavior of banks.
Rime (2001) finds support for Rochet`s second approach, which found that capital requirements are effective in an incomplete market. Specifically, he examines whether and how Swiss banks react to any constraints regarding their capital. He argues that the capital requirements of Switzerland may better reflect the long-term effects because Switzerland has more experience with risk-based capital requirements than the United States. Moreover, he uses an empirical approach instead of a
theoretical. Over a period of six years, he finds that Swiss banks tend to increase their ratio of capital if they are close to the minimum level of the requirement, which supports the motivation for regulatory pressure. However, Rime (2001) states that the level of risk is not adjusted under regulatory pressure. Thus, it is questionable to what extent his results hold if capital requirements are adjusted because a bank may
increase its risk-taking due to a change in the capital requirement.
In addition to why capital requirements can reduce excessive risk-taking, Morrison and White (2005) mention the skin in the game argument. If banks do not have enough equity at stake, they make suboptimal investment decisions from the society’s point of view. This is because it is commonly assumed in banking literature that banks act in the interest of shareholders. Consequently, it is possible that banks take excessive risk at the expense of debt holders. In other words, if banks do not put their own skin in the game, or hold enough capital, it is possible that they increase risk-taking. Again, this is a moral hazard problem. Therefore, imposing a minimum
capital requirement can help to reduce excessive risk-taking.
Morrison and White (2005) also discuss the ability of a regulator to detect a bank’s risk-taking. If a regulator’s ability to detect banks that take excessive risk is high, the reputation is considered to be good. In case of a good reputation, their theory states that capital requirements can prevent moral hazard problems. This statement can only hold if it is assumed that no crises will occur. To see why, the assumption of this statement of Morrison and White (2005) is motivated by the notion that if public confidence in a regulator detecting bad banks is sufficiently high, crises will not occur. Therefore, they suggest that regulators with a good reputation follow a loose capital requirement policy. Ceteris paribus, regulators with a poor reputation should always follow a tighter capital requirement policy since it is possible that a crisis may occur. To conclude, capital requirements can help to reduce moral hazard problems depending on the regulator’s reputation.
Laeven and Levine (2009) examine the relationship between capital requirements and ownership structure of a bank in order to control a bank`s risky behavior. In an empirical setting, they find that capital requirements can either increase or decrease a bank`s risk-taking depending on its ownership structure. The relationship between capital requirements and risk-taking is positive if the bank has a sufficiently powerful owner. On the other hand, when there are many owners the relationship is negative. Thus, any change in capital requirements will have different effects on banks depending on their ownership structure. Furthermore, Laeven and Levine (2009) argue that this issue matters from a public policy point of view. In other words, a supervisor should take the ownership structures of banks into account when the policy with respect to capital requirements is set. According to them, ignoring ownership structure leads to an incomplete analysis of capital regulations on a bank`s risky behavior. However, they do not discuss possible solutions to deal with this issue. Their main result does provide a useful insight in helping to answer the research question.
To conclude, several factors of risk have been discussed. Moral hazard holds a positive relationship with risk-taking while franchise value holds a negative one. Different views exist on the relationship of competition with risk-taking. In order to answer the research question, it is important to recognize how these factors interact with risk and perhaps take them into account. Graph 1 at the end of this section tries to plot the degree of each factor against the level of risk-taking based. Note that the
graph is simplified and tries to illustrate what has been discussed before. Then, a discussion on imposing capital requirements showed that it can motivate to reduce a bank’s risk-taking. This thesis seeks to answer the research question by using an empirical approach. To the knowledge of this paper, little empirical literature exists on the effect of approaching the minimum capital requirement on European Union banks. Therefore, an important contribution of this thesis is to assess this effect empirically. The methodology of this thesis is based on Rime’s methodology (2001) because it examines the effects of approaching the minimum capital requirement. However, this thesis partly follows Rime’s methodology and uses a simplified model. Over the period 2000-2013, observations across 129 banks are made. A panel data regression allows to make implications about the effectiveness of capital
requirements. For example, if it is found that approaching the minimum capital requirement is significant in the regression model with a negative coefficient, then they contribute to decrease a bank’s risk-taking. Moreover, the minimum capital requirement would justify the goal of a supervisor, which is to ensure the safety and soundness of the banking system by reducing risk-taking.
Graph 1. Relationships with factors of risk-taking
DE G R E E OF F A C T OR RISK-TAKING
Moral Hazard Franchise value Competition
Methodology
Earlier it was discussed that there are two ways to respond to meeting the minimum capital requirement. Either by increasing capital or decreasing risk-taking. This thesis focuses on decreasing risk-taking and the model is based on instrumental variable regression. The methodology of this thesis is partly based on Rime’s (2001). However, he uses a different model with respect to the effect of approaching the minimum capital requirement on reducing risk-taking. This model is outside the scope of this thesis. Therefore, the model is simplified to instrumental variable regression. To the knowledge of this thesis, statistical inferences from the model of this thesis can still be made because econometric properties are respected as explained in the data analysis section. This thesis also recognizes that improvements to the model can still be made.
To capture the effect of approaching the minimum capital requirement, it will be referred to as regulatory pressure. As mentioned earlier, this thesis defines
regulatory pressure as banks who are either in 0.4% or 0.2% range of the minimum capital requirement. Note that this range is arbitrary and can be adjusted if needed. As the introduction briefly discussed, the European Parliament has agreed that the
minimum capital requirement is set at 8% of total capital to RWA. Thus, banks from the dataset are subject to regulatory pressure if their ratio of total capital to RWA is between 7.8% and 8.2% or between 7.9% and 8.1% in the other regression. Rime (2001) uses a probabilistic approach to define regulatory pressure. For simplicity, this thesis uses a dummy variable to define regulatory pressure. That is, the variable is sufficient if a bank’s capital adequacy ratio is in the range of 7.8%-8.2%, or 7.9%-8.1%, and zero otherwise.
The equation looks as following: 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷_𝐷𝐷1𝑗𝑗,𝑡𝑡 = 𝛽𝛽0+ 𝛽𝛽1𝐷𝐷𝑅𝑅𝑅𝑅 + 𝛽𝛽2𝐿𝐿𝐿𝐿𝐿𝐿𝐷𝐷𝐷𝐷𝑗𝑗,𝑡𝑡+ 𝛽𝛽3𝐿𝐿𝐷𝐷𝐷𝐷𝑆𝑆𝑅𝑅𝑗𝑗,𝑡𝑡+ 𝛽𝛽4𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗,𝑡𝑡+ 𝑌𝑌𝑅𝑅𝐷𝐷𝐷𝐷𝑡𝑡+ 𝜉𝜉𝑗𝑗,𝑡𝑡. In this equation, the dependent variable is
the change in risk-taking (drisk_d1) while the other variables are independent and defined throughout the remainder of this paragraph. The change in risk-taking is the dependent variable because this thesis focuses on reducing risk-taking when banks approach the minimum capital requirement. The construction of the dependent variable is described in a later paragraph. To start defining and describing the remaining variables, the change in total capital (dtcap) is considered to be an endogenous variable. It has been discussed before that a bank has two ways to
respond to a minimum capital requirement. Either it increases capital or it reduces risk. If it increases capital, the nominator of the capital adequacy ratio is affected. If it decreases risk-taking, the denominator of the capital adequacy ratio is affected. This implies that a change in total capital affects risk-taking through RWA and vice versa. Therefore, the change in total capital is considered to be endogenous. The remainder of this paragraph defines the remaining four exogenous variables. First, Rime (2001) states that bank’s size (lsize) has a relationship with risk diversification. Note that size is defined as total assets and the natural logarithm is used in the regression
model. However, he does not specify whether this relationship is positive or negative. Second, current loan losses (loss) interact with a bank`s riskiness through the ratio of RWA to total assets. According to Rime (2001), an increase in loss on loan provisions leads to a decrease in RWA. Intuitively, an increase in current loan losses results in a lower franchise value. In turn, this increases the risky activities of banks as it was discussed that relationship between franchise value and riskiness is negative. Rime (2001) notes that RWA reflects banks’ decisions on risk-taking with appropriate timeliness. Taken the intuition behind current loan losses and the reflection of RWA together, an increase in the loss on loan provisions leads to a decrease in RWA. Third, regulatory pressure (reg) is taken into account. Last, table 6 in the appendix shows that the regression model should take time-fixed effects into account.
Rime (2001) found that regulatory pressure has an insignificant effect on reducing risk-taking for Swiss banks. Based on this result, one might expect that regulatory pressure does not have a significant effect on reducing risk-taking when the minimum capital requirement is approached. However, the literature review showed that capital requirements can contribute to decreasing risk-taking and the desired outcome of reducing risk-taking can depend on ownership structure, the reputation of the regulator and transparency of the market. Therefore, future research on these factors is needed to fully develop a hypothesis on approaching the minimum capital requirement.
Earlier it was discussed that several banks have failed the comprehensive test and need to adjust their risk-taking. As described in the introduction, the
comprehensive test tests whether banks hold a sufficient level of capital under two scenario. However, the effects of approaching the minimum capital requirement has not been tested empirically. Therefore, the data for the regression will be obtained from the database Bureau Van Dijk because it includes all banks subject to the
comprehensive test except for one. The dataset can be summarized as following. It contains 129 banks and the time series are from 2000 until 2013. It differs from Rime (2001) in the respect that includes European Union banks while Rime (2001)
regresses on Swiss banks. It should be noted that the panel dataset is unbalanced which means that not every value for an independent variable is observed. This might result in estimation errors which will be discussed into more detail in the discussion section.
It should be noted that Rime (2001) defined two capital ratios. First, there is the ratio of capital to total assets (RCTA). Second, there is the ratio of capital to risk-weighted assets (RCWA). The regression model of this thesis is based on RCWA as the regulatory pressure variable measures the ratio of total capital to RWA.
In order to generate the change in risk taking variable, the risk taking variable is first divided by size, which results in the risk ratio variable. In the dataset, risk taking is measured by “Fitch Core Adjusted Weighted Risks” which can be regarded as RWA. In other words, the risk ratio shows the percentage value of RWA to total assets. Next, the first difference of the risk ratio is taken to create the change in risk taking. Furthermore, the risk ratio provides intuition in terms of a bank’s policy with respect to risky activities. If most of the values of the risk ratio show that it is greater than 0.5, then it might be implied that most banks tend to hold mostly assets that are considered to be less risky. Therefore, it might be implied that banks do not take excessive risk if the risk ratio is greater than 0.5. Also, the change in total capital variable is constructed by taking the first difference of total capital.
Furthermore, the lagged variables of the change in total capital and total capital are created in Stata.
To estimate the coefficients of the instrumental variable regression, fixed effects will be taken into account. By definition, fixed effects allow to investigate the causes of changes within a bank because it controls for time-invariant characteristics between banks. In other words, the effects of changes in total capital, meeting the minimum capital requirement, losses on loan provisions, and total assets cannot be biased with respect to omitted time-invariant variables. The instrumental variables used are the lagged variables of total capital and the change in total capital.
Data analysis
First, an analysis on descriptive statistics can be made. Figure 1 shows the percentage change in risk taking over the time period. From the figure it can be seen that there are four outliers. Those four outliers belong to two banks. These two banks have shown a great change in terms of risk-taking between 2005 and 2012. In the regression results, these outliers are not dropped. For most banks the percentage change in risk-taking fluctuates around the value 0 which is displayed on the y-axis. Figure 1.Change in risk
Figure 2 shows the change in total capital in euros. Again, there are two banks to which outliers belong. The outliers in this case exceed the change in capital over the value of 2000 euros, either negative or positive. Also, these outliers are not dropped. Earlier it was discussed why the change in total capital is taken as an endogenous variable. Figure 1 and Figure 2 motivate this as both the change in risk and the change in total capital show a similar trend which could imply that both variables affect each other.
Figure 2. Change in total capital
Table 1 at the end of this section summarizes the independent variables. The values for total capital, the change in total capital and loss in loan provisions are
denoted in € million while the values of the change in risk are denoted in percentage points. Furthermore, the natural logarithm of size, or total assets, is taken. First, it can be seen that the values of all independent variables vary over time because the within variation for each variable does not equal zero. Also, the variables differ across banks as the between variation for each variable does not equal zero. The banks in the European Union held relatively approximately 9800 € million in capital on average, while the average loss on loan provisions is approximately 6% of total capital. The variation of total capital within a bank is relatively small as compared to the variation between banks over 13 years. Consequently, estimation difficulties might arise. To explain why the within variation of total capital is smaller, the size might give intuition. The between variation of the size variable is nearly the same as the overall variation, which indicates that the sample includes both small and large banks. Second, a dummy variable was used to measure regulatory pressure. Table 1
demonstrates that a relatively small amount of banks were in this range as the mean is approximately 0.06 which means that not many banks experienced regulatory
pressure. Third, Table 1 also shows that the change in risk variable is not observed on eighteen banks, observations for total capital are missing for eleven banks, eight banks have no values reported for the regulatory pressure variable, and two banks have no value reported for losses on loan provisions. This relates to the
aforementioned unbalanced panel dataset. These missing values might impact the estimation of the coefficients of the independent variables.
Table 4 in the appendix shows the summary of the risk ratio. As discussed before, this variable represents the ratio of RWA to total assets. The mean is approximately equal to 0.5 which could mean that most banks tend to be balanced between risky activities and non-risky activities. In other words, it implies that they do not necessarily take excessive risk. However, sovereign debt is not included in RWA while it is not riskless. Therefore, the risk ratio might underestimate the representation of risk-taking. The results section will discuss which variables have a significant influence on a bank’s change in risk-taking and whether these impacts are positive or negative.
Table 5 in the appendix shows the result for testing for heteroscedasticity in a fixed-effects model. The null hypothesis is that the model has homoscedasticity. Since the F-statistic is significant, it can be concluded that heteroscedasticity is present in the model. Controlling for heteroscedasticity in the fixed-effects regression model is
necessary in order to estimate the coefficients in the risk equation. Kiefer and Vogelsang (2005) refer to this type of estimation as heteroscedasticity and
autocorrelation constant (HAC). They mention that bootstrapping provides a way to take HAC testing into account. Therefore, this thesis will bootstrap the risk equation in Stata in order to have valid standard errors.
Table 6 in the appendix shows the F-statistic for testing for time-fixed effects. The null hypothesis is that the coefficients for all years are jointly equal to zero. Since the result is significant, the null hypothesis is rejected. Thus, time-fixed effects need to be taken into account. In other words, there are unobserved variables that are the same across all banks but differ over time.
In conclusion, the data analysis has tried to show why the change in total capital is used as an endogenous variable in the regression model. Also, it has been shown that heteroscedasticity is present. By bootstrapping the risk equation the standard errors are valid. Furthermore, time-fixed effects are included in the model.
Table 1. Summarizing variables
Mean Std. Dev. Min Max Observations
Drisk_d1 overall between within -0.0112 0.0792 0.0241 0.0761 -0.7164 -.01111 -0.7238 0.7141 0.0749 0.7246 N=777 n=111 Reg overall between within 0.0649 0.2465 0.1376 0.2221 0 0 -0.685 1 1 0.9935 N=1047 n=121 Dtcap overall between within 380.5221 3109.14 1192.47 2932.021 -40660 -2483.4 -49773.23 35792 9503.75 26678.77 N=791 n=118 Tcap overall between within 9815.033 17948.62 19536.24 3236.453 -3937 13 -23955.17 188357 156920.2 41251.83 N=928 n=118 Lsize overall between within 10.7244 1.756 1.655 0.452 4.4067 6.377 6.290 14.7665 14.4281 13.2947 N=1300 n=129 Lloss overall between within 612.5918 1709.384 1336.523 1157.287 -823 -749.3333 -7645.783 25225 8258.375 17579.22 N=1237 n=127 Results
Table 2 and 3 show the results for the instrumental variable regression. As it was discussed earlier, the change in total capital has been used as an endogenous variable. Its corresponding lagged variable and the lagged variable of total capital were used as instrumental variables. The significance level is set at 5 percentage points. As can be seen from the output from table 2, regulatory pressure is
insignificant in both tables. This is in line with the results of Rime (2001) who found the same outcome using a probabilistic measure for regulatory pressure. This might be counterintuitive because it was discussed that the goal of a supervisor was to ensure the safety and soundness of the banking system. The regression results imply that banks, approaching the minimum capital requirement, do not significantly reduce risk-taking. The discussion section tries to explain this.
Size is significant which means that if a bank increases its total assets by one percentage point, it increases its exposure to risk faced by its portfolio by 4.71 percentage points. This could indicate that when a bank increases its total assets, it decreases its ratio of RWA to total assets. Note that the remaining variables, except for 2006, are insignificant and do not affect the risk that a bank faces.
Table 2. IV regression based on a range of 0.4% for regulatory pressure
Drisk_d1 Observed coef. P-value
Reg** .0144 0.156 Dtcap 9.30e-07 0.774 Lsize -.0471* 0.034 Lloss -2.51e-06 0.569 2003 -.0080 0.752 2004 -.0014 0.946 2005 .0343 0.141 2006 .0538* 0.049 2007 .0282 0.209 2008 -.0102 0.699 2009 .0337 0.157 2010 .0218 0.372 2011 .0161 0.503 2012 .0103 0.669 2013 .0252 0.317 constant .499 0.033 R-squared within 0.0872
*represents significance at the 0.05 level
** regulatory pressure is sufficient when the capital adequacy ratio is between 7.8% and 8.2%
Table 3. IV regression based on a range of 0.2% for regulatory pressure
Drisk_d1 Observed coef. P-value
Reg** .013 0.434 Dtcap 8.53e-07 0.7 Lsize -.046* 0.031 Lloss -1.88e-06 0.597 2003 -.005 0.846 2004 -.017 0.470 2005 .034 0.165 2006 .052* 0.050 2007 .026 0.260 2008 -.014 0.596 2009 .031 0.204 2010 .018 0.451 2011 .012 0.590 2012 .005 0.826 2013 .021 0.403 Constant .489 0.028 R-Squared within 0.0837
*represents statistical significance at the 0.05 level
** regulatory pressure is sufficient when the capital adequacy ratio is between 7.9% and 8.1%
Discussion
The results have indicated that regulatory pressure does not reduce a bank’s riskiness. This might be counterintuitive because the goal of a supervisor was to ensure a safe and sound banking system. In order to do so, one of the ways for a supervisor is to impose a minimum capital requirement to reduce risk-taking. However, banks approaching the minimum capital requirement have not changed their exposure to risk as the insignificant value indicates that the share of RWA in a bank’s portfolio remained the same. Therefore, regulatory pressure does not
significantly affect European Union banks, subject to the comprehensive test, to adjust risk-taking.
In comparison with the results of Rime (2001), he does not find significance for regulatory pressure in the risk equation. Also, he does find that size has a
significant influence but he finds a positive estimated coefficient. Last, he finds that all years are significant, while the regression results of this thesis find that one year is significant. As discussed earlier, Rime’s methodology differs from this thesis as he uses a simultaneous equation while this thesis regresses one equation using
instrumental variables. Consequently, interaction between the change in total capital and the change in risk taking might not have been fully captured which could explain why the results differ. Another explanation could be that Rime (2001) uses a
probabilistic measure to define regulatory pressure while this thesis uses a dummy variable to define it. Furthermore, outliers were not dropped which could have impacted the estimation of the coefficients. More importantly, this thesis has used a simplified regression model. Therefore, it is possible that not all factors have been taken into account. For example, the value of the within R-squared indicates that the explanatory variables explain 8.72 percentage points of all movements of the change in risk taking for each bank individually. Moreover, the literature review has shown that moral hazard, franchise value, and competition are factors that hold a relationship with risk-taking, but these are not taken into account in the regression model. Last, the range of regulatory pressure was determined arbitrarily. The results of table 2 and 3 are nearly the same while different ranges were used. This implies that regulatory pressure with a maximum range of 0.4% does not significantly affect risk-taking. It should be clear that the instrumental variable regression model does not take every factor into account and improvements could be made. However, to the knowledge of
this thesis, this simplified model can be used as a valid regression to estimate coefficients.
Conclusion
This thesis has tried to examine to what extent approaching the minimum capital requirement incentivize banks, who have been tested by the ECB, to reduce their exposure to risk faced by their portfolio from the period 2000-2013. Two ranges, 0.4% and 0.2%, were used to define regulatory pressure and the results are nearly the same. This indicates that regulatory pressure with a maximum range of 0.4% does not significantly affect European Union banks, subject to the comprehensive test, to adjust their level of taking. One way for a supervisor to reduce excessive risk-taking is to impose a minimum capital requirement. Based on these results, one might question the effectiveness of regulatory pressure on banks approaching the minimum capital requirement. However, it should be noted that the regression model was
simplified and other factors have not been taken into account. Therefore, the results of the regression model should not be taken too literally. Rather, it suggests that
improvements could be made by adding more factors of risk-taking such as moral hazard, franchise value, and competition. It also suggests that regulatory pressure does not have the desired outcome for European Union banks, close to meeting the minimum capital requirement.
Bibliography
Berger, A. N., Herring, R. J., & Szegö, G. P. (1995). The role of capital in financial institutions. Journal of Banking & Finance, 19(3), 393-430. Blum, Jürg. (1999). Do capital adequacy requirements reduce risks in banking? Journal of Banking & Finance, 23(5), 755-771.
Beck, T., Demirgüç-Kunt, A., & Levine R. (2006). Bank Concentration, Competition and Crises: First Results. Journal of Banking & Finance,
30(5), 1581-1603.
Boyd, J. H., De Nicoló, G. (2005). The theory of Bank Risk Taking, and Competition Revisited. Journal of Finance, 60(3), 1329-1342.
Demsetz, R. S., Saidenberg, M. R., & Strahan, P. E. (1996). Banks with Something to Lose: The Disciplinary Role of Franchise Value. Economic Policy
Review, 2(2), 1-14.
Keeley, M. C. (1990). Deposit Insurance, Risk, and Market Power in Banking. American Economic Review, 80(5), 1183-1200.
ECB: Aggregate Report on the Comprehensive Assessment. (2014, October). The European Central Bank. Retrieved
from: https://www.ecb.europa.eu/ssm/assessment/html/index.nl.html
Hellmann T.F., Murdock K.C., & Stiglitz J.E. (2000). Liberalization, Moral Hazard in Banking, and Prudential Regulation: Are Capital Requirements
Enough? American Economic Review, 90(1), 147-178.
Laeven, L., Levine. R. (2009). Bank governance, regulation and risk taking.
Journal of Financial Economics, 93(2), 259-275.
Kiefer, N.M., & Vogelsang, T.J. (2005). A New Asymptotic Theory for
Heteroskedasticity-Autocorrelation Robust Tests. Econometric Theory, 21(6), 1130-1164
Morrison, A. D., & White, L. (2005). Crises and Capital Requirements in Banking.
The American Economic Review, 95(5),1548-1572.
Rime, B. (2001). Capital Requirements and Bank Behavior: Empirical Evidence for Switzerland. Journal of Banking & Finance 25(4), 789-805.
Rochet, J. (1992). Capital Requirements and the Behavior of Commercial Banks. European Economic Review, 36(5), 1137-1178.
Appendix
Table 4. Risk ratio variable
Variable Mean Std. Dev. Min Max Observations drisk overall between within .5195 .2264 .2216 .0946 .0058 .0085 .0831 1.519 1.078 1.046 N = 894 n = 111 T-bar = 8.054 Table 5. Heteroscedasticity
Modified Wald test for groupwise heteroskedasticity in fixed effect regression model H0: sigma (i) ^2 = sigma^2 for all i
Prob>chi2 = 0.0000*
* represents statistical significance at the 0.05 level
Table 6. Time-fixed effects Testing for time-fixed effects
H0: coefficients for all years are jointly equal to zero ( 1) 2002.year = 0 ( 2) 2003.year = 0 ( 3) 2004.year = 0 ( 4) 2005.year = 0 ( 5) 2006.year = 0 ( 6) 2007.year = 0 ( 7) 2008.year = 0 ( 8) 2009.year = 0 ( 9) 2010.year = 0 (10) 2011.year = 0 (11) 2012.year = 0 (12) 2013.year = 0 Prob > F = 0.00000*
* represents statistical significance at the 0.05 level
