Bachelor Thesis
Your name: Dikran Bolte
Your student number: 10354638
Date: 02/02/2016
Specialization (within Economics and Business): Economics and Finance
Field: Finance
Number of credits thesis: 12ECTS
Title of your thesis: The effect of capital ratios on beta in the banking system
The effect of capital ratios on beta in the banking system D.A.P. Bolte
10354638 02/02/16
BSc Economics and Business
Statement of Originality
This document is written by Student Dikran Bolte who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
Lately a lot of emphasis in the debate about the stability of the banking system has been put on increasing capital. Banks argue that capital ratios are sufficiently high while governments, the media and regulators rationalize higher required capital ratios. The efficiency of the regulations that are currently implemented is however not broadly discussed. By estimating the effect of CET1 ratios of global banks on their respective CAPM-beta it is found that these two variables have a positive relation implying that the current regulatory framework is not totally efficient.
Introduction
The Common Equity Tier 1 (CET1) ratio has recently been a point of discussion with the introduction of Basel III, the new global regulatory framework for more resilient banks and banking systems. The Basel Committee introduced their view on bank capital in 1988 by splitting up total capital in multiple tiers. The Tier 1 or core capital is the key element of total capital according to the Basel Committee. Basel I, the first minimum capital requirements came into law in 1992, followed by Basel II in 2008. In the aftermath of the financial crisis regulators have put an emphasis on higher capital requirements along with other regulations, to be fully implemented in 2019 as Basel III. Banks have actively lobbied to counteract the implications of increased capital requirements. Previous research has not come to consent over whether increasing capital requirements decreases bank risk and stability. Some research implicated diminishing risk factors when capital requirements and ratios increased (Furlong
and Keeley, 1989, p. 891), others found the opposite to be true (Blum, 1999, p. 768). However, not much emphasis is put on what the effects are on bank stability due to higher CET1 ratios specifically.
Common Equity is made up of common stock and retained earnings of the respective bank (Basel Committee on Banking Supervision, 2010, p. 13). This is the capital that can be dissolved in case of a bankruptcy or a bail-in. To derive the CET1 ratio common equity is divided by the risk weighted assets (RWA). RWA consists of the different assets a bank holds, weighted to its respective risk class. In essence the CET1 ratio reflects a bank’s ability to absorb losses and its limit (ceteris paribus) of the amount of debt capital that can be invested into risky financial activities.
In order to hold the CET1 ratio stable while investing in assets with a positive risk weight, banks can either issue new common shares or retain their earnings. Both of these measures are more expensive in times of economic downturn than during booms. Furlong and Keeley (1989, p. 888) state that value-maximizing banks will, in order to expand their asset base, raise more capital to hold their capital ratios stable. They found no positive relation between increased capital ratios and asset risk, quantified as the standard deviation of the rate of return on assets. Another way to keep the CET1 ratio constant while taking on more assets is to invest in assets with a lower risk weight. Banks keeping their CET1 ratios stable will therefore have a procyclical effect on the lending behavior and the return stability of the banking system as a whole (Basle Committee on Banking Supervision, 1999, p. 16).
Currently the required CET1 ratio is between 2,0 and 4,5 percent (dependent on whether Basel III is already being implemented in the respective country). A higher CET1 ratio implies a smaller range of possible profits and losses that the respective bank can realize and a less volatile return pattern is expected. On the other hand, Blum (1999, p. 768) states that higher capital ratios might actually increase risk. This is due to the fact that a higher
capital ratio reduces future expected profits. Lower expected future profits will decrease the incentive to avoid default and will encourage capital arbitraging, adjusting a bank’s assets in order to be able to place them in a lower risk class without actually reducing economic risk (Jones, 2000, p. 36). From this might be inferred that RWA do not perfectly, or even adequately, describe the risk weight of a bank’s assets. In that case, the expectation that a higher CET1 ratio will induce a less volatile return pattern may prove to be unfounded.
In order to evaluate bank stability I will focus on the relative correlation of the return of banking stocks with the market return. This will enable detection of banks’ exposure to risks (Altunbas, Manganelli & Marques-Ibanez, 2011, p. 20). The relative correlation of the stock return with the market return is captured in the beta (𝛽! =!"#(!!, !!)
!"#(!!) ) of the Capital
Asset Pricing Model (CAPM): 𝑟! = 𝑟!+ 𝛽!∗ (𝐸 𝑟! − 𝑟!) and is an important determinant of the risk premium required by investors (Demarzo, 2007, p. 401). Since a higher CET1 ratio entails a relatively higher portion of equity to RWA this will lower the chance of bankruptcy and concurrently narrow the range of possible loss and profit outcomes thus bringing beta closer to zero.
Fama and French (2004, p. 44) state that the CAPM might not be applicable in many situations due to its strong underlying assumptions. Mainly the assumption of rational investors that only care about the mean and variance of the returns of a portfolio in the next period is viewed as unrealistic (Fama & French, 2004, p. 37). However, for the goal of this paper there is enough comfort to use the CAPM, mainly because the beta values are calculated from actual returns and expected returns are not incorporated in the model. In this case the actual beta values are expected to possess enough explanatory strength about the correlated variance to be of significant value. A possible limitation of using the CAPM in this research is the difficulty for investors to observe banks’ CET1 ratios real-time since they are
not reported real-time. One could argue that this rejects the assumption of a perfectly efficient equity market.
Investing in more risky assets requires holding more common equity than when investing in less risky assets retaining a constant CET1 ratio. Although a different risk profile of investments might require a different financing structure, it is assumed that this does not affect the outcome of this research. According to Modigliani and Miller (1958, p. 271) regardless of how a firm is financed (by equity or debt) the value of the firm is the same. Although there will be no emphasis on market capitalization in this research, the possible argument that raising equity instead of debt would be more costly is found to be untrue. However, financially banks are unique companies since their assets are mainly debt products and this assumption might not be applicable to banks (Miller, 1995, p. 485). This paper will not focus on whether the Modigliani and Miller theorem holds for financial institutions.
Although there is agreement on the increased size of equity held by banks as a result of capital regulation, there are no conclusive results on the implications of risk based capital regulation (VanHoose, 2007, p. 3694). Since the CET1 ratio is risk weighted, an increase in the CET1 ratio could actually drive banks to invest in more risky assets or find other ways, like capital arbitraging, to earn a high return on equity that might induce more risks and interfere with the stability of the banking system. Most research focuses on the effects of increased capital requirements on the capital held by banks and their lending behavior. However, there is little research with a focus on the link of higher capital ratios and the stability of performance.
Other factors like ownership structure (Laeven & Levine, 2009, p. 273), regulations (Kim & Santomero, 1988, p. 1231) and competition (Beck, 2008, pp. 21-22) could directly or indirectly, through the degree of possible managerial moral hazard, influence risk taking by
banks and bank performance stability. However, these issues are beyond the scope of this research.
The model
The goal of this paper is to examine how CET1 ratios affect bank performance stability, quantified by the beta value of the CAPM. The following variables are included in the model: value of beta, CET1 ratio, Tier 2 capital ratio, total assets (in billions of Euros), net interest margin (NIM) and a dummy for the financial crisis. These variables are briefly discussed in order to provide intuitive understanding of the model and endorse its application.
The beta from the CAPM is a measure for the correlated covariance of required return on equity with the required return for the market. In other words the beta is the riskiness that is related to the stock, the less risky the stock is the more stable the performance of the respective bank. The beta is calculated by dividing the covariance of the weekly changes in dividend-adjusted stock prices and the weekly changes in dividend-adjusted market prices (S&P500) by the variance of the weekly changes in dividend-adjusted market prices. The S&P500 is chosen as the market proxy because of the increased comparability between banks betas and the generalizability of empirical results. Benchmarking errors will be limited since the vast majority of banks in the sample operate internationally and the stock and market prices are adjusted for dividends (Roll, 1980, p.9).
A higher CET1 ratio would entail relatively more capital to absorb negative events such as an economic downturn or an unexpected high portion of distressed loans. Concurrently the amount of funds that can be set out as loans and invest in other opportunities are limited without taking other actions because more investment would entail a rise in RWA and a decrease in the CET1 ratio. So a higher CET1 ratio decreases the opportunities to
realize returns. Limited availability of funds to realize returns and a decreased chance of going bankrupt will result in a smaller range of possible loss and profit outcomes.
Another effect stemming from lower CET1 ratio is the increased presence of moral hazard. When a bank has low capitalization shareholders are more likely to take risks since they have less to lose in the case of bankruptcy or a bail-in. Although regulations do require a minimum level of equity, there are instances where the expected payoff will be higher for shareholders when deviating to more risky profiles (VanHoose, 2007, p. 3688). Since shareholders only make a profit if interest costs of liabilities to creditors are paid, shareholders have higher incentives to deviate to more risky assets in bad times when these interest costs are less likely to be fully paid. A higher CET1 ratio reduces the moral hazard since there is more equity at stake for shareholders. In conclusion, a higher CET1 ratio is expected to result in a lower value of beta.
The Tier 2 capital ratio and the CET1 ratio together form the total capital ratio. Although Tier 2 capital is lower quality capital than CET1 capital it may cause the classification of a different risk profile for the concerned stock. A higher portion of Tier 2 capital will reduce a banks risk profile even if the bank holds this subordinated capital and reserves on top of its CET1 capital. The Tier 2 capital ratio is expected to affect the beta value negatively because of a lower risk profile resulting in a lower correlated variance. However, since Tier 2 capital is lower quality capital than CET1 capital the effects of the Tier 2 capital ratio are expected to have weaker effects on the value of beta than the CET1 ratio has.
The amount of assets could influence the CET1 ratio. Larger banks might hold lower levels of CET1 because they are more diversified, have relatively fewer risk-weighted assets and expect to need less excess reserves. On the other hand, smaller banks are under less prudent inspection because they have less systematical importance (Boyd & Runkle, 1993, pp. 48-49). This can cause a moral hazard, which will result in holding lower levels of CET1.
Chen and Hsieh (1985, p. 456) constructed twenty portfolios entailing different sized companies and performed a time-series regression on their betas. By testing the estimated residuals they found that smaller firms are more risky than larger firms (Chen & Hsieh, 1985, p.470). However, Boyd and Runkle (1993, p. 63) find no prove for the relation of asset size and risk of bankruptcy. This might prove the relation between asset size and riskiness for financial institutions to be unfounded. Since the amount of total assets could reduce both beta and CET1 ratios this would influence expected results. Although the evidence is not conclusive, a larger asset base is expected to result in a less risky firm quantified by a lower value of beta.
The net interest margin is included as a measure of profitability and reflects the interest spread between a banks assets and liabilities divided by its average earning assets. One strong determinant of the NIM is risk aversion (Ho & Saunders, 1981, p. 598). Since a lower NIM shows a smaller relative interest spread it entails more risk averse behavior and would decrease the correlated variance of the required return, the NIM would be expected to be positively related to the regressand. More risky assets earn a higher interest margin increasing the NIM and this would be expected to increase the value of beta. Concurrently this would increase the RWA, which in turn reduces CET1. However, by using assets more efficiently a bank can also realize a higher NIM, not affecting the CET1 ratio but affecting beta nonetheless. The respective bank might in general be more capable of allocating its assets more efficiently, decreasing its maximal losses and thus reduce the value of beta. In conclusion, there is no strong expectation on the effect of the NIM on the value of beta since the underlying reason for the level of the NIM will be the determinant of the effect.
Finally, a dummy variable for the financial crisis (2008-2009) is included in order to account for the abnormal period within the sample range, which according to Berger and
Bouwman (2013, p. 155) may have implications for the expected negative relation of the CET1 ratio and the value of beta.
The sample consists of data from 157 of the 2000 largest listed banks1. The banks are registered in 33 different countries on all continents except for Africa. Realized betas are calculated from weekly data retrieved from Yahoo Finance. All other data are retrieved from Bankscope. In summary, the main model that will be estimated is the following:
𝛽! = 𝑐 + 𝛼!𝐶𝐸𝑇1 𝑟𝑎𝑡𝑖𝑜 + 𝛼!𝑇2 𝑟𝑎𝑡𝑖𝑜 + 𝛼!𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 + 𝛼!𝑁𝐼𝑀 + 𝛼!𝑓𝑐 + 𝜀!
For a summary and a graphical oversight of the data, refer to Figure 1 and Table 1 and 2. Figure 1 shows the average values of the beta measure and capital ratios2. Table 1 shows the summary statistics on the variables included in the model. Table 2 shows the correlation between the variables.
1 Measured by total assets and selected on availability of sufficient data. Sufficient data is
quantified as the presence of data on the main variables for at least three years in the sample.
2 The Tier 2 ratio being the difference between the Total capital ratio and the CET1 ratio. 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 2006 2007 2008 2009 2010 2011 2012 2013 2014 A ve rage c ap ital r ati os A ve rage b eta
Figure 1. Trend of average values
Methodology
After assigning an integer to every individual bank, the values of the beta are taken. This enables the estimation of a linear model where a higher value of the beta stands for a higher correlated variance. All observations with a negative beta value or a beta value of more than 4,0 are dropped since they are regarded as outliers or measuring mistakes. First, CET1 ratio is regressed against the beta value to estimate the model using ordinary least squares (OLS) with
Table 1. Summary statistics on the regression variables
Variable Min Max Mean Standard dev. Number of obs.
Beta (β) 0,001 3,190 0,802 0,549 1162
CET1 ratio 4,20 28,60 10,98 3,18 948
Tier 2 ratio 0,00 9,90 2,92 1,53 945
Total assets (€bln) 1,09 2770 176 373 1069
Net interest margin -0,03 25,96 2,97 2,26 1056
Table 2. Correlations of the regression variables
Variable Beta CET1 ratio Tier 2 ratio Total assets NIM Financial crisis
Beta 1,0 CET1 ratio -0,0190 1,0 Tier 2 ratio 0,0402 -0,4102 1,0 Total assets 0,3185 -0,0320 -0,0929 1,0 NIM -0,1359 0,3145 -0,0555 -0,2168 1,0 Financial crisis 0,0315 -0,0729 0,0748 -0,0002 0,0454 1,0
robust standard errors. Second, the control variables are independently incorporated to observe their individual effects on the estimation of the model. To control for omitted variable bias there will be an OLS regression done with fixed country effects. This estimation will be carried out in the same way as the multiple linear OLS regression.
In order to discern which effects the CET1 ratio has on the bank-specific correlated variance with the market, a fixed effects panel regression model with robust standard errors of the unbalanced sample is estimated. Here as well all control variables are incorporated independently to observe their individual effects on the estimation of the model. Fixed effects are used after performing a Hausman-test on the complete model. The null-hypothesis that a random-effects model adequately estimates the coefficients of the model is resoundingly rejected (chi2=30,363).
On all the coefficients of the variables a t-test is carried out to see if they significantly differ from zero. The focus is on the t-test outcome of the CET1 ratio. If the null-hypothesis (H0: α1 = 0; H1: α1 ≠ 0) is rejected, this test is powerful in developing a case for the influence
of the CET1 ratio on bank performance stability.
Empirical results
In this section the results will be discussed. First, the outcomes of the estimation using OLS (with and without fixed effects) and the panel regression will be reported. Second, the estimations of the fixed effects OLS and the panel regression will be analyzed and possible causes of the outcomes are discussed. Finally, the differences of the two estimation outcomes will be analyzed. To provide some graphic insight in the simple linear OLS regression of the CET1 ratio against the value of beta, refer to Figure 2. Although Figure 2 is a great simplification of the estimated multivariable model, it does strengthen the expectation of a
negative relation between the CET1 ratio and beta. For quantitative results, refer to Table 3 (OLS), 4 (OLS with fixed country effects) and 5 (panel).
Table 3 shows that although a downward sloping regression line is estimated, the results for the CET1 ratio are not significantly different from zero. Adding the control variables independently does not increase the significance of the coefficient of the CET1 ratio. Also, the sign of the estimated coefficient for the CET1 ratio turns positive, which opposes expectations. Furthermore it shows that the effect of the Tier 2 ratio is significantly different from zero on a five percent level and the total assets and the NIM are significantly different from zero on a one percent level. This implies that the Tier 2 ratio and total assets have a different relation to beta than expected. The NIM is estimated to have a negative relation with beta, supporting the expectation of a higher NIM due to increased efficiency. The estimated coefficient for the financial crisis is not significantly different from zero. The signs of the estimated coefficients, except for the financial crisis dummy, are inconsistent with expectations. 0 0.5 1 1.5 2 2.5 3 3.5 4 0.00 5.00 10.00 15.00 20.00 25.00 30.00 Be ta CET1 ratio
Controlling for omitted variable bias by including country fixed effects has an impact on the significance of the estimated coefficients (see Table 4); only the coefficient for the financial crisis dummy is not significantly different from zero. However, also in this estimated model most coefficient signs are inconsistent with expectations: a percentage point increase of the CET1 ratio of 1,0 is estimated to have a 0,0265 percentage point effect on beta. This would imply that banks with higher CET1 ratios have a higher correlated variance. This effect seems small regarding the standard deviation of the value of beta of 0,549. However, the range of CET1 ratios in this sample is over 24 percentage points and the standard deviation is 3,18. Although marginal effects are small, actual observed effects might be substantial.
Table 3. Linear OLS regression with robust standard errors
Variables Coefficient (t-value) Coefficient (t-value) Coefficient (t-value) Coefficient (t-value) Coefficient (t-value) CET1 ratio -0,0042 -0,0048 0,0042 0,0089 0,0094 (-0,77) (-0,07) (0,66) (1,30) (1,37) Tier 2 ratio 0,0140 0,0290** 0,0308** 0,0302** (0,93) (2,01) (2,14) (2,09) Total assets 0,0005*** 0,0004*** 0,0004*** (8,01) (7,52) (7,52)
Net interest margin -0,0194** -0,0200***
(-2,56) (-2,64) Financial crisis 0,0461 (1,13) Constant 0,8500*** 0,7678*** 0,5862*** 0,5907*** 0,5808*** (13,89) (7,55) (5,94) (5,97) (5,87) Number of obs. 948 945 945 938 938 R2 0,0006 0,0016 0,1064 0,1123 0,1134 * significant at 10% level ** significant at 5% level *** significant at 1% level
The positive relation between the CET1 ratio might be attributable to different causes: i) banks with high CET1 ratios have a higher beta because they could easily lower this ratio, increasing the range of possible profit and loss outcomes, without any regulatory restrictions. This would increase the correlated variance since investors do not only observe the range of profit and loss outcomes related to the current CET1 ratio but also the range of profit and loss ranges that can be easily realized by decreasing the CET1 ratio. ii) If banks with high capital ratios tend to have these because of capital arbitrage and this would be observable by investors, the beta value might be higher because not all risks are incorporated in the RWA. However, if these assumptions were true regulators could observe this as well and would at least be expected to introduce a possible solution or iii) beta is determined by more than only
Table 4. Linear OLS regression with fixed country effects and robust standard errors
Variables Coefficient (t-value) Coefficient (t-value) Coefficient (t-value) Coefficient (t-value) Coefficient (t-value) CET1 ratio 0,0118* 0,0244*** 0,0235*** 0,0265*** 0,0265*** (1,78) (3,30) (3,15) (3,48) (3,49) Tier 2 ratio 0,0585*** 0,0593*** 0,0576*** 0,0576*** (3,97) (4,04) (3,96) (3,93) Total assets 0,0002** 0,0002** 0,0002** (2,53) (2,28) (2,27)
Net interest margin -0,0466*** -0,0466***
(-3,92) (-3,91) Financial crisis -0,0002 (0,00) Constant 0,2045*** -0,0860 -0,0836 -0,0714 -0,0714 (2,85) (-0,83) (-0,81) (-0,69) (-0,69) Number of obs. 948 945 945 938 938 R2 0,3174 0,3315 0,3379 0,3488 0,3488 * significant at 10% level ** significant at 5% level *** significant at 1% level
the included variables in the model. This would imply that the estimation of the model still suffers from omitted variable bias.
The Tier 2 ratio has a higher positive impact on the value of beta than the CET1 ratio (F-test: 5,414). Banks with higher CET1 ratios tend to have lower Tier 2 ratios. When
regressing the CET1 ratio against the Tier 2 ratio with fixed country effects and robust standard errors a coefficient of -0,7171 is estimated (t-value: -9,345). This can be a result of a managerial or shareholder preference for a balanced (leveraged) financing structure or Tier 2 capital might be less needed because the CET1 ratio is sufficient to comply with regulations. Value maximizing banks do not want to go bankrupt while maximizing profits. Holding a higher Tier 2 capital ratio, while not necessary with respect to total capital ratio requirements, might infer the presence of risks not incorporated in RWA. This will have a stronger effect on the value of beta than the CET1 ratio because of the higher risk associated with the extra portion of not incorporated RWA. The reasons why higher capital ratios increase the value of beta will most likely (partly) overlap.
Contrary to expectations the coefficient for total assets has a positive sign implying that larger banks have a higher value of beta. This effect is rather small but acknowledging the large range (2769 billion Euro) and standard deviation (373 billion Euro) the impact can prove to be substantial. Although Chen and Hsieh (1985, p. 470) found smaller firms to be more risky this might not be the case with financial institutions. Bigger banks might be more tempted and more able to undertake capital arbitrage on a larger scale, implying the presence of risks that are not incorporated in the RWA. Also the presence of moral hazard in large banks compared to small banks might stimulate capital arbitrage, where this might be different for firms in other industries. Since this research was not focused on firm size and
4 Significant on a 5% level 5 Significant on a 1% level
total assets was merely incorporated in the model as a control variable, these results are not generalizable.
The NIM is estimated to have a negative effect, significantly different from zero, on the value of beta. This would imply a link between the NIM and efficient asset usage or debt financing structure. More efficient asset usage would boost interest income while a more efficient debt financing structure would reduce interest cost, both increasing the NIM and reducing the maximal losses. Since the level of efficient asset usage or financing structure is beyond the scope of this research, there is no additional quantitative evidence to be put forward on this possible explanation. Also, the CET1 ratio and the NIM are positively correlated. This is most likely caused by the decreased interest cost related to relative more equity funding.
Although a negative relation between the financial crisis and the value of beta is estimated this result was not significantly different from zero. This could both be caused by the (mis)specification of the financial crisis period. This position is enforced by the results estimated with a regression model with a financial crisis dummy for 2008 only. The estimation was significantly different from zero (t-value: -2,176) and did have a substantial effect (-0,1023) on the value of beta while the other coefficients remained more or less stable. It is also possible that the returns showed no different correlative variance from the market during the whole or part of the financial crisis than in other years.
The estimated regression model on panel data with fixed effects shows a higher resemblance with expectations although the sign of the CET1 ratio is estimated to be positive as well. Only the CET1 ratio and total assets are significant different from zero on a five percent level (see Table 5).
If a bank is to increase its CET1 ratio with 1,0 percentage points its beta will increase with 0,021 percentage points. Since these results are estimated with fixed bank effects there might be different explanations than in the previous OLS regression with fixed country effects. Quantitatively these findings will not have a high marginal impact but could have substantial total effects. It is remarkable that the CET1 ratio has a positive effect on the value of beta within a financial institution. This implies that when a bank increases its CET1 ratio ceteris paribus, it experiences a higher correlative variance with the market. Due to the risk weight incorporated in the CET1 ratio this can most likely be explained by inefficiencies in the risk weighing system, triggering capital arbitrage or insufficient amounts of investment opportunities in certain risk classes. Conclusively, there are multiple possible causes of the
Table 5. Panel regression with fixed effects and robust standard errors
Variables Coefficient (t-value) Coefficient (t-value) Coefficient (t-value) Coefficient (t-value) Coefficient (t-value) CET1 ratio 0,0200** 0,0185 0,0198** 0,0207** 0,0205** (2,17) (1,97) (2,12) (2,32) (2,30) Tier 2 ratio -0,0117 -0,0133 -0,0148 -0,0145 (-0,55) (-0,63) (-0,70) (-0,67) Total assets -0,0003** -0,0003** -0,0003** (-2,48) (-2,43) (-2,38)
Net interest margin -0,0161 -0,0153
(-0,62) (-0,58) Financial crisis -0,0066 (-0,19) Constant 0,5862*** 0,6364*** 0,6782*** 0,7193*** 0,7194*** (5,83) (4,89) (5,06) (4,71) (4,70) Number of obs. 948 945 945 938 938 R2 0,0006 0,0007 0,0762 0,0614 0,0633 * significant at 10% level ** significant at 5% level *** significant at 1% level
positive relation of the CET1 ratio and the beta value: i) the Basel Committee did not assign adequate risk weights to different investing opportunities. Banks, knowingly or not, make use of this to increase the riskiness of their assets without having to adjust its financing structure to keep the CET1 ratio constant. Over time banks tend to improve their capital arbitraging techniques, resulting in a higher CET1 ratio without lower actual risks or ii) if a bank increases its CET1 ratio this might be the case because it does not have enough investing opportunities in a by shareholders desired risk class or it expects to need more common equity to account for possible negative shocks. This would implicitly increase the possible profit and loss outcomes. The basis of both explanations is inefficiency of the risk weighing system for assets and the possible exploitation performed by banks of this system. Another possible explanation is the presence of reverse causality. Banks with higher betas might be more tempted to retain a higher CET1 ratio due to the bigger range of possible profit and loss outcomes. Although this would imply a misspecification of the model, there is currently no strong scientific literature that indicates this.
The estimated coefficient for total assets is estimated to be negative, which supports the earlier expectation that a larger asset base reduces risk. When a bank increases its assets it has at least some investment opportunities it sees as profitable whereas when it decreases assets it might not have sufficient profitable investment opportunities. This might explain the relation to the value of beta. It is also feasible that banks with a larger asset base are more diversified, both in terms of activities and geography. However, due to the setup of the model and the limited availability of data this effect is not clearly observable.
No significant effect is estimated of the Tier 2 ratio on the value of beta. Investors might not view the Tier 2 ratio as an important variable for the determinacy of the correlated variance and therefor the result is insignificant. This is the most likely reason for the Tier 2 ratio to have no estimated effect on the value of beta within a bank.
If the increase of the NIM originates from a larger portion of equity funding, resulting in lower interest costs but higher financing costs a negative relation to the value of beta is explained by the smaller range of possible profit and loss outcomes. If the increase of the NIM were attributable to overall efficiency this would result in less negative profit outcomes. The range of possible profit and loss outcomes would become smaller since the minimal values would become higher, causing a lower value of beta.
On the contrary, an increase in the NIM due to higher interest income (and increased risk) would most likely increase the value of beta. It is most likely that this is not the overall cause of the increase of the NIM in most instances. However, again due to the setup of the model and the limited availability of data the cause is not clearly discernible.
The possible misspecification of the financial crisis period could also in the panel regression be the cause of the insignificant estimation of the financial crisis dummy coefficient.
The estimations from the panel regression and the OLS regression with fixed country effects yielded similar estimation outcomes. The CET1 ratio is estimated to have a positive relation to the value of beta in both regressions. The most viable cause of the positive relation between the CET1 ratio and the value of beta in both regressions is that the CET1 ratio is sensitive to whether all actual risks are incorporated in RWA through the risk weighing system. The gap between my expectations and the estimation results imply that the risk weighing system applied to calculate RWA is in that sense not fully efficient and might be sensitive to capital arbitrage. The CET1 ratio is established as a capital requirement to enhance the banking system and make it more stable. However, by using capital arbitrage banks riskiness and possible loss and profit outcomes are not accurately reflected by the CET1 ratio. This effect
might be so large that a higher CET1 ratio is mostly present in banks with a high value of beta.
The introduction of the leverage ratio introduced in Basel III, which is not adjusted for asset risk, might be a partial solution but is also sensitive to off-balance sheet banking. In this regard, more or different regulations might be needed to ensure more trustworthiness in a stable banking system. The same applies to the Tier 2 ratio that is also closely related to the determination of the amount of RWA.
Another possible explanation of the positive relation between the CET1 ratio and the value of beta is reversed causality. Since regulations on the CET1 ratio restrict banks from lowering the ratio to zero banks are not fully flexible in setting their CET1 ratios. In downturns market returns decrease and this decrease is relatively higher for banks with a higher value of beta. Concurrently, they are unable to lower their CET1 ratio below a certain threshold because of restrictions put in place by regulators. Banks with a higher value of beta will retain a higher CET1 ratio on average in order to not be obliged to raise more CET1 capital or reduce RWA in downturns. This would imply reversed causality.
Concerning the difference in the asset coefficient estimations in both regressions, this can most likely be explained by a misspecification of the model. Asset size might not have a linear relation to the value of beta. Although bigger banks have a higher correlated variance in general, a further increase overall lowers the value of beta.
Conclusion
Contrary to expectations the CET1 ratio is estimated to have a positive relation to the value of beta. This implies that a higher CET1 ratio actually reduces bank stability. The most likely cause of these findings is the inefficiency of the current asset risk weighing system. Since some banks are able to implement capital arbitrage better than others or increase their ability
over time the CET1 ratio might have a large signaling effect on investors. The CET1 ratio might be a good tool to increase bank stability but could probably be implemented in a more effective way, limiting capital arbitrage and increasing the true informative value on its purpose.
Another possible explanation is reversed causality. This would entail that banks with higher values of beta will hold higher CET1 ratios in order to not be restricted during downturns.
The outcomes of this research imply that most banks do not implement their optimal asset size. While this research was not intended to assess this issue, the findings were unambiguous.
Due to the estimated relation of the NIM to the value of beta it can be concluded that an increase of the NIM in general stems from a relative increase of equity funding or increased asset efficiency. An increase of the NIM due to higher interest income would not decrease the value of beta.
Although the financial crisis did not seem to have any significant effect on bank stability this could nonetheless be relevant because of a misspecification of the financial crisis dummy variable. A different specification of the financial crisis dummy variable might be of more informative value and have a significant effect on the value of beta.
In order to improve the stability of the banking system more attention should be paid to the realized implications of existing regulations. Further research could be done on different capital requirements or other regulatory variables and different definitions of bank stability should be considered.
References
Altunbas, Y., Manganelli, S. & Marques-Ibanez, D. (2011). Bank Risk During the Financial Crisis (Working Paper no. 1394). Retrieved from European Central Bank website: https://www.ecb.europa.eu/pub/pdf/scpwps/ecbwp1394.pdf
Basel Committee on Banking Supervision. (1999). Capital Requirements and Bank
Behaviour: the Impact of the Basle Accord. Retrieved from: http://www.bdresearch
.org/home/attachments/article/485/bcbs_capital%20requirement%20and%20effect.pdf Basel Committee on Banking Supervision. (2010). Basel III: A Global Regulatory
Framework for more resilient banks and banking systems. Retrieved from:
http://www.bis.org/publ/bcbs189.htm
Beck, T. (2008). Bank Competition and Financial Stability: Friend or Foes? (Working Paper no. 4656). Retrieved from: http://papers.ssrn.com/sol3/papers.cfm?abstract
_id=1152483
Berger, A. & Bouwman, C. (2013). How does capital affect bank performance during financial crises? Journal of Financial Economics, 109, pp. 146-176.
Blum, J. (1999). Do Capital Adequacy Requirements Reduce Risks in Banking? Journal of
Banking and Finance, 23, pp. 755-771.
Boyd, J. & Runkle, D. (1993). Size and Performance of Banking Firms. Journal of Monetary
Economics, 31, pp. 47-67.
Chen, N. & Hsieh, A. (1985). An Exploratory Investigation of the Firm Size Effect. Journal
of Financial Economics, 14, pp. 451-471.
Demarzo, B. (2007/2011/2014). Corporate Finance. Boston, MA: Pearson Education. Fama, E. & French, K. (2004). The Capital Asset Pricing Model: Theory and Evidence.
Journal of Economic Perspectives, 18(3), pp. 25-46.
Furlong, F. & Keely, M. (1989). Capital Regulation and Bank Risk-Taking: A Note. Journal
of Banking and Finance, 13, pp. 883-891.
Ho, T. & Saunders, A. (1981). The Determinants of Bank Interest Margins: Theory and Empirical Evidence. The Journal of Financial and Quantitative Analysis, 16(4), pp. 581-600.
Ioannidis, C., Pasiouras, F. & Zopounidis, C. (2010). Assessing Bank Soundness with Classification Techniques.
Jones, D. (2000). Emerging Problems with the Basel Capital Accord: Regulatory Capital Arbitrage and Related Issues. Journal of Banking and Finance, 24, pp. 35-58. Kim, D. & Santomero, A. (1988). Risk in Banking and Capital Regulation. The Journal of
Laeven, L. & Levine, R. (2009). Bank Governance, Regulation and Risk Taking. Journal of
Financial Economics, 93, pp. 259-275.
Miller, M. (1995). Do the M & M propositions apply to banks? Journal of Banking &
Finance, 19, pp. 483-489.
Modigliani, F. & Miller, M. (1958). The Cost of Capital, Corporation Finance and the Theory of Investment. The American Economic Review, 48(3), pp. 261-297.
Roll, R. (1980). Performance Evaluation and Benchmark Errors (I). The Journal of Portfolio
Management, 6(4), pp. 5-12.
VanHoose, D. (2007). Theories of Bank Behavior under Capital Regulation. Journal of
