Excess capital of European banks: does bank heterogeneity matter? University of Limoges, LAPE, 5 rue Félix Eboué BP3127, 87031 Limoges, France First Draft: February 2009 This version: May 2009 Preliminary version Abstract: Using a sample of European banks over the period 1992-2006, we show that factors such as cost of equity, risk, size, peer and market discipline can explain banks excess capital but that these general results conceal important differences across banks: the determinants of excess capital differ depending on banks size, activity and funding mode. More precisely, we show that the market discipline, economic cycle, and ex post risk variables are significant only for banks heavily relying on market fundings. This suggests that market exerts a pressure on those banks so that they hold more excess capital. We also show that market pressure influences both large and small banks, but that the excess capital of small banks appears influenced by the regulatory pressure. Thus, given the high heterogeneity of excess capital behaviours highlighted in this study, we can suspect that the consequences of Basel II through its first pillar will be contrasted. For example, banks highly relying on market fundings could be only slightly affected by the refinement of the capital minimum adequacy ratio (Pillar 1) in Basel II if market pressure is what matters most for these banks.
1. Introduction After the creation of the Basel committee in 1974 and more precisely since 1985, nearly all the Basel Committee countries have placed substantial reliance on specific capital ratios calculations that were increasingly based on a risk-weighting of assets. However, it was not until 1988 that a formal minimum capital requirement was firstly introduced at the international level through what became known as Basel I *. The main motivation of Basel I was the risk posed to the stability of the global financial system by low capital levels of international active banks and by the phenomenon known as "the race to the bottom" that is "one country's lower regulatory standards make it more difficult for other countries to maintain rigorous but necessarily more costly standards". Hence, under Basel I, to meet the prudential regulation guidelines, banks must have Tier 1 capital of 4% of risk weighted assets (RWA) and total regulatory capital (Tier 1 + Tier 2) at least equal at 8% of RWA. The main aim of Basel I was therefore to raise capital ratios of international active banks and so it seems to have done (Ediz et al. (1998), Rime (2001) for Europe, Jacques and Nigro (1997), Aggarwal and Jacques (1997), Hancock and Wilcox (1997) for US). Since the beginning of 1990s, an upward trend in capital ratios has been noticed throughout G-10 countries and has aroused interest in questioning why banks hold such high capital ratios, or put differently, why they hold capital in excess of what is required by the regulator. Throughout this paper, excess capital is defined as the difference between the actual capital ratio ((Tier1+Tier2)/Risk weighted assets) and the Basel minimum required capital ratio (8%), except for special cases. For instance, the Financial Stability Authority (FSA) sets two separate capital requirements for each UK bank: a trigger ratio, which is the minimum individual capital ratio; and a target ratio set above the trigger. We therefore follow Jokipii and Milne (2008) and consider 9% minimum capital requirement ratio for all UK banks. The rationale of this inquiry stems from the fact that bankers often argue that an excessively high capital ratio reduces their ability to compete in their activities because capital * The final agreement was signed on July 11, 1988. The capital standard became effective in March 1989 and internationally active banks were required to achieve the benchmark by December 1992. For Japanese banks, the deadline was March 1993, the end of their accounting year (Ito and Sasaki (2002)). Tarullo (2008), p.53. Several modifications subsequently intervened, the most notable being the 1996 amendment to incorporate market risk beyond the already existing credit risk in the minimum requirement. 2
is more expensive than debt. Many academics therefore expect that banks operate close to their required minimum capital ratios. Different studies have tried to determine what the determinants of excess capital are (Lindquist, 2004, Nier and Bauman, 2006, Jokipii and Mine, 2008) or how banks adjust their capital ratios (Wall and Peterson, 1995, Berger et al., 2008, Flannery and Ragan, 2008). Several factors have been identified: internal factors, regulatory pressure, competition, market pressure, and cyclical behavior. Most studies consider internal factors such as the cost of holding capital, the size of the bank, its charter value or its risk taking. Dietrich and Vollmer [2005] demonstrate theoretically, using a Nash bargaining model, that excess capital does not serve as a buffer against shocks to project returns, but as a strategic tool for renegotiations with borrowers. In the same vein, Bernauer and Koubi [2006] show both theoretically and empirically that better capitalized banks experience lower borrowing costs. Thus, banks excess capital can be viewed as a mean to compete with other banks. Excess capital may also be explained by market pressure. Nier and Baumann [2006] examine how market discipline can influence banks capital excess formation. They mainly show that, ceteris paribus, stronger market discipline, taken into account via the portion of uninsured liabilities and information disclosure, leads to a higher bank capital excess. Finally, there is a strand of literature that focuses on capital cyclical behavior. Jokipii and Milne [2008] and Ayuso, Perez and Saurina [2004] ** show, for example, that the economic cycle (captured through the GDP) and excess capital are negatively related. In this paper, we consider banks from 15 European countries over the period 1992-2006. Our set of banks discloses an average excess capital of 7.04. In other words, European banks hold on average a capital ratio of 15% over the period studied, which is almost two times the Basel minimum required capital. Besides, we denote a high volatility of this excess capital as the standard deviation is 5.58. This volatility may suggest heterogeneity across banks and/or across time. Hence, the main novelty of this paper lies in the fact that we take into account banks characteristics. While most of the previous studies consider all banks equally, or focus only on banking specialisation (commercial, savings and cooperative banks (Jokipii and Milne (2008)), we split our sample into six categories, two by two, depending on They use a substantial cross-country panel data of 32 countries from 1993 to 2000. ** Jokipii and Milne [2008] use a sample of European banks from 1997 to 2004 and Ayuso, Perez and Saurina, [2004] consider Spanish banks from 1986 to 2000. Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal, Spain, Sweden, Switzerland, and United Kingdom. 3
three factors: bank activity type, financing mode and size. Indeed, we suspect that the determinants of excess capital could differ depending on these characteristics. Thus, the aim of this study is twofold: we investigate the determinants of European excess capital and why it is so heterogeneous. With this aim in view, we consider a large set of potential explanatory factors including different proxies of capital regulation and market discipline, and control variables identified in the previous literature. Our hope is that, in such a way, we can deduce the potential impact of the minimum capital adequacy revision underway, namely Basel II. We suspect that a potential modification in the risk weight, which could result in the reduction of the amount of the capital required by regulation, should be felt differently according to the degree of exposure to the market. For instance, all else being equal, the higher capital in excess a given bank holds before Basel II due to the market pressure, the less this bank is expected to change its capital position after Basel II. We also add to the above previous literature in the sense that, by taking into account economic cycle, competition, risk-taking, we are able to provide comparable results with studies which had looked at these elements separately using different samples. The rest of the paper is structured as follows. In section 2, we define the sample of banks and describe the set of variables. Section 3 is dedicated to a statistical analysis. We present the method used to estimate the relationship between excess capital and the explanatory variables in section 4 before presenting the results in section 5. Section 6 concludes. 2. Sample and variables 2.1 Sample The sample consists of 442 banks from 15 European countries over the period 1992-2006. We retain banks for which information about the total capital ratio is available in Bankscope Fitch IBCA at least 7 years on the period under study. Appendix 1 presents the distribution of banks by country and specialization. Total capital ratio is (Tier 1 + Tier 2)/ Risk weighted assets and is used to construct our dependent variable. 4
2.2. Dependent variable The dependent variable corresponds to what is commonly called buffer in the literature that is the amount of capital banks hold in excess of what is required by national regulators. More precisely, we construct the variable excessk as the bank s total Tier 1 plus Tier 2 capital Basel 1 risk-weighted capital ratio less its regulatory minimum requirements. This regulatory minimum requirement is set to 8% in most countries of our sample except in Germany where it is set to 12.5% for newly established banks in the first two years of business (Caprio et al. (2008)) and in the United Kingdom where we consider 9% (Jokipii and Milne (2008)). We delete observations for which the capital ratio is either negative or in the top five percent of the sample distribution as such high capital ratios might distort our findings. 2.2. Independent variables We consider a large set of variables likely to affect banks excess capital. These variables are defined in Table 1. Data come from Bankscope Fitch IBCA and Datastream. This paper draws on previous works done in Europe and elsewhere that have questioned why banks do hold capital in excess of what requires the prudential regulation. Several factors have been identified: internal factors inherent to the bank itself and external factors such as: regulatory, supervisory and market pressure, and economic cycle. Most studies have in common internal factors such as the cost of holding capital, the size of the bank, its charter value or its risk taking (Alfon et al. (2003), Heid et al. (2004), Memmel and Raupach (2007), Kleff and Weber (2008), Gropp and Heider (2008), Flannery and Ragan (2008), Berger et al. (2008)). Drawing on these studies and others, we describe hereafter the expected relationship between the dependent (excessk) and exogenous variables: We consider the fact that the capital in excess could simply reflect an unusual period of bank profitability (Flannery and Rangan (2008)). When raising new capital is costly, capital accumulation could rely on internally generated funds, in line with the Pecking order theory of finance. Bankers may increase capitalization through higher retaining 5
earnings and weaker dividend payments and stock repurchase. We therefore expect a positive sign between profit and capital in excess. In a world different from that of Modigliani and Miller (1958), equity is more costly compare to other bank liabilities because of information asymmetries. Equity may also be disadvantaged because interest payments on debt are deducted from earnings before tax. Excess capital is hence expected to be negatively associated with this cost. Direct measurement of this cost is difficult. Therefore, previous studies have considered the return on equity (ROE) as a proxy variable for the direct cost of remunerating excess capital. By the way, as stressed by Jokipii and Milne (2008), ROE reflects both cost and revenue. To eliminate the revenue side already taken into account through profit, we orthogonalize the variable ROE. Indeed, as we denote a high and significant correlation between the variables ROE and profit, we regress ROE on the profit variable. The variable roe used in the regressions corresponds to the residuals of this estimation that is the part of ROE which is not explained by profit. Thus, we expect a negative relationship between roe and excess capital. Since a bank s probability of failure depends on its risk profile and its capital level, we expect that good bank management implies that the more the risk it plans to take, the more the capital it keeps aside. Nevertheless, as argued in Nier and Baumann (2006), this positive relationship applies only for ex-ante measures of risk, proxied here by rwaa. Realized (or ex-post) risk are rather found to reduce capital ratios. Ex-post measures used in the literature are non performing loans or loan loss provisions over total assets (Ayuso et al. (2004)). Due to data availability, we use the latter in this study. Thus, we expect a negative relationship between llpa and the dependent variable and a positive one when we consider rwaa, proxy for ex-ante risk, assuming prudent management. All else equal, an increase in assets through loan granted should increase the capital requirements ratio and therefore decrease the excess capital (Ayuso et al. (2004). Thus, we expect a negative relationship between loang and the dependent variable. 6
The level of capital should be influenced by regulatory requirements. Therefore, we use a capital regulatory index capreg, proposed by Barth et al. (2004), which is a summary index calculated from initial capital stringency and overall capital stringency. Higher levels of capital stringency are expected to be positively linked with the capital in excess that banks hold as greater stringency should induce prudent behaviour (Schaeck and Cihak (2007)). We also consider that banks, on the one hand, whose customers are more sensitive to default risks, particularly uninsured ones (subordinated debt holders for instance), and on the other hand, which operate in highly competitive environment, are expected to hold more capital than prescribed by the regulator. In the former case, the bank faces what we may call the traditional market discipline. In the latter case, imperfect information makes banks hold high levels of capital as a signal to differentiate themselves from their peers. We name this peer discipline. When we consider traditional market discipline, we suppose that uninsured bondholders and investors may doubt of the capacity to resist bankruptcy when their bank operates with a capital ratio very close to the regulatory minimum. The very possible choice that the bank faces is therefore either to increase its debt remuneration, or hold more capital than required by the regulation. Hence, we suppose that the type of fundings can impact excess capital. It should be positively related to the proportion of market fundings, markliab, because their holders are the creditors who have the most incentives to exert a discipline (Nier and Baumann (2006)). We also consider the potential discipline exerted by shareholders (listed dummy variable) and an index, privmonitor, reflecting the extent of private supervision by country (Barth et al. (2004)). Considering peer discipline, excess capital may serve as an instrument, which the bank is willing to pay for, in the competition with its peers for unsecured deposits and money market funding (Lindquist (2004), Dietrich and Vollmer (2005), Bernauer and Koubi (2006), and Schaeck and Cihak (2007). Thus, we consider the mean excess capital of competitors, compexcessk, which should positively affect excess capital. A consensus among the previous literature also emerges: it indicates that larger banks hold less average capital in excess due to scale economies in screening and monitoring and the diversification effect. The dependent variable should be negatively related to size. Another reason for large banks to hold a smaller buffer may be their Too Big To Fail 7
nature. Thus, any kind of implicit insurance should be associated with lower excess capital. The level of capital banks hold may also depend on macroeconomic conditions in the bank s country of origin. We therefore introduce the economic cycle to determine whether it has any effect on the capital held by institutions. Previous studies have mostly shown that excess capital and economic cycle tend to be negatively linked (Ayuso et al. (2004), Lindquist (2004), Jokipii and Milne (2008)). This is to say that banks tend to decrease their excess capital during the upturn and increase it in the downturn. The rationale for this may be found in Berger et al. (1995) who argue that banks may hold excess capital to be able to exploit unexpected investment opportunities. We therefore expect a negative link between Gross Domestic Product growth, gdpg, and the capital in excess. 8
Table 1: Description of the independent variable Variable Mnemonic Definition Mean Standard deviation Min Max Expected Sign of the coefficient Profitability profit Post tax profit/ Total assets 0.75 0.80-4.98 19.65 + Equity cost roe Return on equity = Net income/ Equity -0.23 6.09-108.11 70.70 - Ex-ante risk rwaa Risk weighted assets and off-balance sheet risks (inferred from the Cooke ratio)/ Total assets 0.67 0.22 0 1.69 + Ex-post risk llpa Loan loss provisions/ Total assets 0.38 0.47-1.67 6.03 - Peer discipline compexcessk Annual mean of the buffer of banks in the same country and of the same specialization 7.41 3.11-4.90 23.70 + Market discipline listed Dummy variable equal to one if the bank is listed, and 0 otherwise 0.15 0 1 + privmonitor Private monitoring index from Barth, Caprio, and Levine (2004) 8. It ranges from 0 to 9, higher value indicating more private supervision. 5.61 1.04 5 8 + markliab (Total money market fundings+ Subordinated debt)/ Total liabilities 11.56 12.37 0 97.13 + Credit demand loang Annual net loan growth 12.88 18.35-98.00 219.26 - Off-balance sheet activities obsa Off- balance sheet/ Total assets 16.29 21.93 0 236.44 + Control variables gdpg Annual growth rate of the gross domestic product (deseasonalized) 2.05 1.47-0.99 15.43 +/- size Logarithm of total assets 14.23 2.34 9.18 21.17 - Regulation capreg Capital regulatory index from Barth, Caprio, and Levine (2004) 9. It ranges from 0 to 9, with a higher value indicating greater stringency. 6.13 1.04 2 8 + 8 See appendix 2. 9 See appendix 3. 9
3. Statistical analysis In a first step, we conduct a statistical analysis. Accounting data for individual banks are obtained from Bankscope Fitch IBCA. Table 2 shows descriptive statistics for the sample of banks. Table 2: Summary accounting statistics from 1992 to 2006. Excessk Total Assets ( th.) Total Market Fundings/ Total Liabilities (%) Total Deposits/ Total Assets (%) Total Net Loans/ Total Assets (%) Mean 7.04 24038900 11.56 64.47 58.57 Median 5.40 1193400 7.43 64.49 59.91 Maximum 26.10 1.57E+09 97.13 97.88 99.65 Minimum -7.90 9750 0 0.00 0.01 Std. Dev. 5.58 93852231 12.37 17.77 18.42 Total market fundings corresponds to (Total money market fundings + Subordinated debt). Total money market fundings includes certificates of deposit, commercial paper, debt securities, and securities loaned. Net loans are: gross loans loan loss reserves. We can notice that the sample of banks exhibits a relatively high level of heterogeneity. The excess capital is on average 7.04 and we denote a high volatility of this excess as the standard deviation is 5.58. This volatility may suggest heterogeneity across banks and/or across time. Besides, we can see that the sample of banks presents high differences in terms of size, activity type, and financing mode. Thus, we decide to further investigate the link between these variables and excess capital. Figure 1 depicts the annual means of excess capital both for the whole sample of banks and for different sub-groups over time. These sub-groups are constructed depending on the size of the bank, its activity type and its financing mode. We can notice that the annual average of excess capital increases across the period: it starts from 4.15 % in 1992 and finishes at 6.09% in 2006. 10
When we consider the different sub-groups, we can see that large banks hold a lower excess capital than small ones during the whole period. Considering the type of activity, banks with a high ratio of net loans/ total assets and those with a small one have similar excess capital at the beginning of the period but their excess capital sharply differ later, banks heavily involved in credit activity having smaller one. We find opposite results when we consider the financing mode, we see that banks heavily relying on deposits have lower excess capital than banks with a low ratio of deposits/ total assets until 2003, then we can see that their excess capital converge. In order to confirm statistically these differences in terms of excess capital on the different sub-groups, we run difference of means tests. Table 4: Annual excess capital means of sub-groups: difference of means tests Means HIGH LOW Total assets 4.65 9.68 44.93*** Deposits/ Total assets 6.63 7.75 7.47*** Net loans/ Total assets 4.58 9.81 22.17*** Significance test for difference in means between HIGH and LOW (t-stats) Total assets is considered as high if it is greater than one billion Euros and low if it is less than one billion Euros. The ratio of deposits/total assets is considered as high if it is greater than its mean value (64.47) plus half a standard deviation (17.77) and low if it is less than its mean value minus half a standard deviation. The ratio of net loans/total assets is considered as high if it is greater than its mean value (58.57) plus half a standard deviation (18.42) and low if it is less than its mean value minus half a standard deviation. The results of these tests confirm that annual excess capital means are significantly different between the sub-groups. Thus, we suspect that the determinants of banks buffer may be different depending on their size, their type of activity and their funding mode. Henceforth, we decide to investigate it econometrically. 11
Sub-sample by size 12 11 10 9 8 7 6 5 4 3 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 Small banks Whole sample of banks Large banks 11 10 Sub-sample by activity type 9 8 7 6 5 4 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 Banks with low (NL/TA) ratio Banks with high (NL/TA) ratio Whole sample of banks Sub-Sample by funding mode 10 9 8 7 6 5 4 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 Banks with low (D/TA) ratio Banks with high (D/TA) ratio Whole sample of banks Figure 1 : Trends of excess capital for 15 European countries between 1992 and 2006 depending on banks size, activity type and financing mode. Total assets is considered as high if it is greater than one billion Euros and low if it is less than one billion Euros. The ratio of deposits/total assets is considered as high if it is greater than its mean value (64.47) plus half a standard deviation (17.77) and low if it is less than its mean value minus half a standard deviation. The ratio of net loans/total assets is considered as high if it is greater than its mean value (58.57) plus half a standard deviation (18.42) and low if it is less than its mean value minus half a standard deviation. 12
4. Method Our model is defined in Eq. (1). Subscripts i and t denote bank and period respectively. excessk = α + α profit + α roe + α llpa + α rwaa + α compexcessk i, t 0, i 1 i, t 1 2 i, t 1 3 i, t 1 4 i, t 1 5 i, t 1 + α privmonitor + α markliab + α loang + α gdpg + α size + α capreg + α listed 6 i 7 i, t 1 8 i, t 9 i, t 10 i, t 11 i 12 i, t + α obsa + u Eq(1) 13 i, t 1 i, t As we introduce two indexes (privmonitor and capreg) that are different across countries but constant on the period, it is not possible to apply fixed effects. We test the relevance of the random effects specification applying the Breusch and Pagan (1980) Lagrange multiplier test which tests if the variance of the random component is zero. The hypothesis of no random effect is clearly rejected that is why we apply a random effect approach. Following the previous literature (Lindquist (2004), Nier and Baumann (2006) for instance), we introduce one year lag in explanatory variables which are susceptible to be endogenous to avoid simultaneity problems. In a first step, this model is estimated on the whole sample of bank. Then, we consider several sub-groups depending on banks financing mode, activity and size. Our intuition is that the behaviour of banks in terms of excess capital may be affected by these characteristics. For example, banks heavily relying on market fundings may be subject to higher pressure from the market than banks relying on insured deposits. We can also suppose that large banks considered as Too big to fail may have less incentives to hold large excess capital. Our sub-samples are defined on the basis of three criteria: - Financing mode: banks heavily (slightly) relying on deposits are defined as those with a ratio deposits/total assets greater (lower) than its mean value (64.47) plus (minus) half a standard deviation (17.77). - Credit activity: banks with a high (low) credit activity are defined as those with a ratio of net loans/ total assets greater (lower) than its mean value (58.57) plus (minus) half a standard deviation (17.42); - Size: banks with total assets greater (lower) than one billion Euros are considered as large (small) banks; 13
5. Results First, we present our empirical results for the whole sample of banks. Then, we take into account sub-samples constructed on the basis of banks financing mode, type of activity and size. We aim to capture the determinants, if any, of the heterogeneity of excess capital depending on these three characteristics. 5.1. General framework Two general remarks emerge from our findings reported in table 5. First, the bulk of our variables are significant with the expected sign most of the time. As hypothesised, banks with high profit have higher capital in excess confirming the pecking order theory. High cost of holding and remunerating equity (roe) has a negative impact on the level of capital in excess. All our indicators of disciplining effect suggest that European banks are in general disciplined by their peers, monitors (rating agencies for instance) and market participants as compexcessk, privmonitor and markliab are highly significant and their coefficients are positive. The loan demand variable loang is, as expected, negatively and significantly related to excess capital indicating that high annual loan growth increases the capital requirement ratio and therefore reduces the bank excess capital. The size coefficient is significantly negative which gives support to the notion of Too Big To Fail status, diversified portfolios and economies of scale in screening and monitoring enjoyed by large banks. Second, contrary to the majority of previous studies (Lindquist (2004), Ayuso et al. (2004), Jokipii and Milne (2008)), bank excess capital and economic cycle are positively linked, a result which is nevertheless in line with that of Bikker and Metzemakers (2007) who offer a cross-country analysis of bank capital pro-cyclicality and find a positive relationship between capital level and business cycle for European countries. Hence, our results suggest that European banks seem to build their cushion during the upturn and to hold less capital excess in the downturn. This could dampen the feared pro-cyclicality problem usually found in the literature because banks could save capital during good times which could be timely needed during bad times. Another unexpected result relates to the ex-ante risk (rwaa) coefficient which is significantly negative. Notwithstandingly, it is in line with findings of many previous researchers who have either fail to find any relationship between ex- ante risk and excess capital (Lindquist (2004)), or a similar significant negative relationship (Bikker and Metzemakers (2007), Jokipii and 14
Milne (2008)). Hence, this result suggests that European banks hold less capital in excess subsequently to the risk taken. This may suggest less precaution toward default risk. Table 5: Determinants of banks excess capital Variable Coefficients profit i,t-1 1.33*** 5.81 roe i,t-1 llpa i,t-1 rwaa i,t-1 compexcessk i,t-1-0.08*** -2.65 0.45* 1.79-4.47*** -7.23 0.29*** 8.75 privmonitor i 0.27** 2.011624 markliab i,t-1 0.04*** 4.94 loang i,t -0.02*** -3.13 gdpg i,t 0.12*** 3.38 size i,t -1.06*** -7.95 capreg i -0.007-0.02 listed i,t 0.02 0.04 obsa i,t-1 constant 0.002 0.64 19.57*** 6.29 Number of observations 1696 R 2 0.44 LM test 0.00 Endogenous variable is excessk i,t. ***, ** and * indicate significance at 1%, 5% and 10% levels, respectively. T- statistics in parentheses are calculated using White heteroscedasticity consistent standard errors. LM test is Breusch and Pagan (1980) Lagrange multiplier test. H0 is no random effects. The p-value > χ2 (1) is reported. 15
To pin down possible factors responsible for the high variability of excess capital among European banks, we present the results for banks grouped by the financing mode, the type of the predominant activity and the size respectively. 5.2. Financing mode Table 6 reports the results obtained when we consider two sub-groups defined on the basis of their financing mode. The striking result is the different behaviour pattern that we notice for almost half of the regressors once we split the initial sample used above. More precisely, five out of thirteen variables enter differently in the model once we consider differences in financing mode. First, the indicator of discipline exercised by uninsured debtholders (markliab) discloses different behaviours. For banks with high proportion of deposits at their balance sheet markliab is not significant whereas it is noticeably significant and positively related to excess capital for banks that finance themselves strongly from the market. This clearly indicates how external participants do impact the way banks behave in choosing their capital structure. All else equal, being turned towards market for fundings and therefore under heightened surveillance, induce banks to choose higher capital in excess. Second, the ex-post measure of risk llpa and the economic cycle variable gpdg are positively and significantly related to excess capital only for banks with less deposit at their balance sheet. Good economic conjuncture and sheltering against realised risk make banks with less deposit hold high excess capital. Third, loan demand loang variable is significant with the expected negative sign only for banks with a low proportion of deposits and the cost of equity roe is negative and significant only for banks with a high proportion of deposits. For the remaining variables used in the regression, they tell quite the same story for both the whole sample and the sub-samples. 16
Table 6: Determinants of banks excess capital by financing mode Variable High (D/TA) ratio Low (D/TA) ratio profit i,t-1 1.59** 2.34 roe i,t-1-0.12*** -2.68 llpa i,t-1-0.65-1.22 rwaa i,t-1-3.99*** -2.63 compexcessk i,t-1 0.24*** 5.61 privmonitor i 0.28 1.06 markliab i,t-1 0.05 0.99 loang i,t -0.01-1.01 gdpg i,t -0.13-1.59 size i,t -0.77*** -6.18 capreg i -0.42-0.83 listed i,t -0.95* -1.70 obsa i,t-1 0.0004 0.05 constant 18.50*** 5.20 1.51*** 3.09-0.05-1.11 0.93** 1.96-3.89*** -5.42 0.44*** 5.98 0.57 1.49 0.04*** 3.22-0.01** -2.12 0.23*** 3.27-1.05*** -5.10 0.23 0.62 1.43* 1.69-0.01-0.82 14.40*** 4.59 Number of observations 264 661 R 2 0.43 0.44 LM test 0.00 0.00 Endogenous variable is excessk i,t. The ratio of deposits/total assets is considered as high if it is greater than its mean value (64.47) plus half a standard deviation (17.77) and low if it is less than its mean value minus half a standard deviation. ***, ** and * indicate significance at 1%, 5% and 10% levels, respectively. T-statistics in parentheses are calculated using White heteroscedasticity consistent standard errors. LM test is Breusch and Pagan (1980) Lagrange multiplier test. H0 is no random effects. The p-value > χ2 (1) is reported. Thus, the difference observed in terms of excess capital between banks heavily relying on deposits and banks more dependent on market fundings could be explained by the relative pressure of the market and by the positive influence of ex post risk and economic cycle on the excess capital of banks heavily relying on market fundings. 17
5.3. Activity type We then look at the type of activity of the bank. Table 7 shows the results obtained when we separate banks on the basis of their type of activity: we discriminate banks according to the importance of their credit activity. Table 7: Determinants of banks excess capital by type of activity Variable High (NL/TA) ratio Low (NL/TA) ratio profit i,t-1 0.86*** 3.72 roe i,t-1 llpa i,t-1 rwaa i,t-1 compexcessk i,t-1-0.05** -2.19 0.63 1.13-0.37-0.44 0.47*** 3.93 privmonitor i 0.38 0.92 markliab i,t-1 0.02 0.96 loang i,t -0.01-0.81 gdpg i,t -0.05-0.33 size i,t -0.19-0.70 capreg i -0.21-0.68 listed i,t -0.46-0.33 obsa i,t-1 constant -0.007-1.27 2.77 0.82 0.73 0.86-0.08* -1.68-0.73-0.78-5.55*** -2.72 0.18** 1.85-0.04-0.06 0.05* 1.76-0.006-0.89-0.006-0.04-1.40*** -0.04-1.28-1.58-0.11-0.06-0.02-1.63 39.20*** 10.74 Number of observations 221 159 R 2 0.32 0.50 LM test 0.00 0.00 Endogenous variable is excessk i,t. The ratio of net loans/total assets is considered as high if it is greater than its mean value (58.57) plus half a standard deviation (18.42) and low if it is less than its mean value minus half a standard deviation. ***, ** and * indicate significance at 1%, 5% and 10% levels, respectively. T-statistics in parentheses are calculated using White heteroscedasticity consistent standard errors. LM test is Breusch and Pagan (1980) Lagrange multiplier test. H0 is no random effects. The p-value > χ2 (1) is reported. 18
As in the previous case, the results are quite different on the two sub-groups. We can notice that, market discipline (markliab) and ex ante risk (rwaa) variables only influence banks with low credit activity whereas the profit variable (profit) has a positive impact only for banks with high credit activity. By contrast, the cost of equity (roe) and the peer discipline (compexessk) seem relevant for the two sub-groups of banks. Thus, even if it is less clear-cut, we find quite the same results, when we consider the financing mode and the type of activity that is banks largely involved in market activities, whatever in the assets side or in the liabilities side, are subject to higher pressure from the market. This could explain why they hold a relatively higher excess capital. 5.4. Size Finally, we separate large banks from small banks. Results in Table 8 show that size matters in the way European banks set their excess capital. However, we can notice that several variables influence similarly both large and small banks' excess capital like roe, compexcessk, and markliab. Thus, contrary to the previous cases, the market pressure is relevant for the two sub-groups of banks. Three variables are relevant only for large banks: the profit variable, llpa, and the private monitoring index. More interestingly, gdpg and capreg are significantly and positively related to excess capital only for small banks. This could partially explain why these banks hold more excess capital than large banks. Capreg is significant at the one percent level; it indicates that the stringency of regulatory capital influences strongly the behaviour of small banks in terms of excess capital. We can suppose that these banks are more concerned by regulatory pressure as, in case of problems, it can be more difficult for them to raise fresh capital. Thus, they can decide to hold excess capital by fear of falling below the required minimum. The significance of gdpg means that small banks increase their buffer during good economic times which suggests prudent behaviour as this cushion will help during hard times when raising new capital is difficult, particularly for small banks. 19
Table 8: Determinants of banks excess capital by banks size Variable Small banks Large banks profit i,t-1 0.49 0.99 roe i,t-1-0.92*** -9.85 llpa i,t-1-0.22-0.44 rwaa i,t-1-11.80*** -5.12 compexcessk i,t-1 0.44*** 3.87 privmonitor i -1.54-1.59 markliab i,t-1 0.05*** 7.15 loang i,t -0.04** -2.38 gdpg i,t 0.19* 1.72 size i,t -0.56*** -2.99 capreg i 4.06*** 4.14 listed i,t 6.11* 1.78 obsa i,t-1 0.03 1.19 constant 0.52 0.10 1.38*** 4.46-0.07** -2.44 0.57** 2.06-3.92*** -8.82 0.20*** 6.34 0.29* 1.92 0.04*** 4.65-0.01** -2.28 0.02 0.44-0.81*** -6.04-0.07-0.27-0.20-0.56 0.0006 0.22 16.08*** 4.91 Number of observations 551 1145 R 2 0.56 0.24 LM test 0.00 0.00 Endogenous variable is excessk i,t. Large banks are those with total assets greater than one billion Euros and small banks are those with total assets less than one billion Euros. ***, ** and * indicate significance at 1%, 5% and 10% levels, respectively. T-statistics in parentheses are calculated using White heteroscedasticity consistent standard errors. LM test is Breusch and Pagan (1980) Lagrange multiplier test. H0 is no random effects. The p- value > χ2(1) is reported. 20
6. Conclusion Why banks hold capital in excess of what is required by prudential regulation is an important investigation, particularly today when academics, practitioners and policy makers are wondering what could be the potential impact of the replacement of the current capital regulation by Basel II. As put by Berger et al. (2008, p. 125) Without understanding bank s observed capital levels and capital cushions, it is hard to predict how they will respond to economic or supervisory changes. Hence, the novel aspect of our paper is to re-set this question in an European context, and to investigate the extent to which the answer depends on the financing mode, activity type and size of the bank. Indeed, with a graphical and statistical analysis we show that excess capital differs significantly depending on these characteristics. When we look at the determinants of excess capital, our results confirm the findings of previous studies (Lindquist (2004), Ayuso et al. (2004), Nier and Baumann (2006), Jokipii and Milne (2008)). All in all, this paper argues that, contrary to the «race at the bottom» phenomenon that prevailed before any formal international capital regulation, banks seem to behave differently. Instead of lowering capital levels in competition, they seem rather willing to hold high capital levels in order to benefit from low debt costs. Indeed, we find that, for all banks, the bank peer competition is a highly significant variable which means that the capital ratio may be perceived as a tool to negotiate low debt cost or to appear in healthy conditions. Other important factors are profit, market discipline, cost of equity, risk, credit demand and size variables. However, we show that these general results conceal differences across banks. When we separate banks according to their size, activity type, and funding mode, we find that the determinants of excess capital are not the same depending on the sub-groups. More precisely, we show that the market discipline, economic cycle, and ex post risk variables are significant only for banks heavily relying on market fundings. This suggests that market exerts a pressure on those banks so that they hold more excess capital. It also implies that these banks hold more excess capital during the economic upturn. According to the size criterium, market pressure influences both large and small banks, but the excess capital of small banks seems to be also influenced by the regulatory pressure and is higher during economic upturn. Thus, given the high heterogeneity of excess capital behaviours highlighted in this study, we can suspect that the consequences of Basel II through its first pillar will be 21
contrasted. As emphasised by Bikker and Metzemakers (2007) It is not clear in advance whether banks will change their capital buffer behaviour after the changeover to Basel II. Indeed, if market participants have their own risk and capital requirements assessment, that is if they do not content with requiring a mere extra fraction of capital beyond the regulatory minimum capital (8%), banks highly relying on market fundings could be only slightly affected by the refinement of the capital minimum adequacy ratio (Pillar 1) in Basel II. 22
References: Aggarwal, R. and K. Jacques, 1997, A Simultaneous Equations Estimation of the Impact of Prompt Corrective Action on Bank Capital and Risk, in Financial Services at the Crossroads: Capital Regulation in the 21st Century, Conference, 26-27 February 1998, Federal Reserve Bank of New York. Alfon, I., Argimon, I., Bascunana-Ambros, P., 2004. What determines how much capital is held by UK banks and building societies? Occasional paper No 22. UK Financial Services Authority. Aschraft, A., 2001, Do tougher bank capital requirements matter? New Evidence from the eighties, Federal reserve bank of New York. Ayuso, J., Perez, D., Saurina, J., 2004, Are capital buffers pro-cyclical? Evidence from Spanish panel data, Journal of Financial Intermediation 13, 249 264. Barth, J. R., G. Caprio, Jr., R. Levine, 2004, Bank regulation and supervision: What works best?, Journal of Financial Intermediation, Vol. 13, pp. 205 248 Berger, A., R. Herring, G. Szegö, 1995, The role of capital in financial institutions, Journal of Banking and Finance, 19, 393-430. Berger, A., R. DeYoung, M. J. Flannery, D. Lee., Ö. Öztekin, 2008, Why Do Large Banking Organizations Manage Their Capital Ratios?, Journal of Financial Services Research 34:123-149. Bernauer, T. and V. Koubi, 2006, Taking Firms and Markets Seriously: A Study on Bank Behavior, Market Discipline, and Regulation, CIS working paper, N 26. Breusch, T., Pagan, A., 1980, The Lagrange multiplier test and its applications to model specification in econometrics, Review of Economic Studies 47, 239 253. Brewer III E., G. Kaufman, L. Wall, 2008, Bank Capital Ratios Across Countries: Why Do They Vary?, Universita Commercial Luigi Bocconi, Paolo Baffi Centre Research Paper Series No. 2008-28. Caprio, G., R. Levine and J. R. Barth, 2008, Bank Regulation and Supervision database (updated June 2008). Dietrich, D., U. Vollmer, 2005, Why do banks hold capital in excess? A functional approach, IWH Discussion papers. Ediz, S., I. Michael, W. Perraudin, 1998, Bank Capital Dynamics and Regulatory Policy, Bank of England. Flannery, M., K. Rangan, 2008, What Caused the Bank Capital Build-up of the 1990s?, Review of Finance 12, 391-429. 23
Gropp, R., and F. Heider, 2008, The determinants of Capital Structure: Some evidence from banks, ZEW Discussion Papers: 08-015. Hancock, D. and J. A Wilcox, 1997, Bank capital, Non-bank Finance, and Real Estate Activity, Journal of Housing Research, 8, pp. 75-105. Heid, F., Porath D., and S. Stolz, 2004, Does capital regulation matter for bank behaviour? Evidence for German savings banks, Deutsche Bundesbank Discussion Paper No 03/2004. Jackson, P., G. Furfine, H. Groeneveld, D. Hancock, D. Jones, W. Perraudin, L. Radecki, M. Yoneyama, 1999, Capital Requirements and Bank Behaviour: The Impact of the Basle Accord, Basel Committee on Banking Supervision Working Paper n.1 (April). Jacques, K. T. and P. Nigro, 1997, Risk-Based Capital, Portfolio Risk and Bank Capital: A Simultaneous Equations Approach, Journal of Economics and Business, 49, pp. 533547. Jokipii, T. and A. Milne, 2008, The cyclical behaviour of European bank capital buffers, Journal of Banking & Finance 32, 1440 1451. Ito, T. and Y. Sasaki, 2002, Impacts of the Basle Capital Standard on Japanese Banks Behavior, Journal of the Japanese and International Economies 16, 372 397. Kleff, V. and M. Weber, 2008, How Do Banks Determine Capital? Evidence from Germany, German Economic Review 9 (3): 354-372. Kwan S. H., 2004, Testing the strong form of market discipline: the effects of public market signals on bank risk, Federal Reserve Bank of San Francisco working paper in Applied Economic Theory 2004-19. Lindquist, K., 2004, Banks buffer capital: how important is risk?, Journal of International Money and Finance 23 (3), 493 513. Memmel, C. and P. Raupach, 2007, How do banks adjust their capital ratios? Evidence from Germany, Deutsche Bundesbank Discussion Paper N 06/2007. Modigliani, F., and M.H. Miller, 1958, The cost of capital, corporate Finance and the theory of investment, American Economic Review 48, 261-297. Nier, E., and U. Baumann, 2006, Market discipline, disclosure and moral hazard in banking, Journal of Financial Intermediation, Vol. 15, pp. 332-361 Repullo, R. and J. Suarez, 2008, The Procyclical Effects of Basel II, CEPR Discussion Papers, DP6862. Rime, B., 2001, Capital requirements and bank behaviour: Empirical evidence for Switzerland, Journal of Banking and Finance, vol. 25 (4), pp. 789-805. Schaeck K. and M. Cihak, 2007, Bank Competition and Capital Ratios, International Monetary Fund working paper 07/216, Washington D. C. 24
Tarullo, K. D., 2008, Banking on Basel. The Future of International Financial Regulation, Peterson Institute for International Economics. 25
APPENDIX 1: Distribution of banks by country and specialization. Country Number Specialization Number Austria 1 Belgium 1 Denmark 2 Finland 6 France 20 Germany 6 Greece 2 Ireland 3 Commercial banks 131 Cooperative banks 222 Investment banks 6 Medium & Long Term Credit Bank Real Estate / Mortgage Bank 9 Savings banks 65 9 Italy 325 Netherlands 10 Portugal 7 Spain 31 Sweden 12 Switzerland 2 United Kingdom 14 Source: Bankscope Fitch IBCA 26
APPENDIX 2: Private monitoring index definition by Barth, Caprio, and Levine (2004) It is based on several criteria: a) Certified Audit Required: This variable is equal to 1 whether an outside license audit is required of the financial statements issued by a bank. Such an audit would presumably indicate the presence or absence of an independent assessment of the accuracy of financial information released to the public. b) Percent of 10 biggest banks rated by international rating agencies: The percentage of the top 10 banks that are rated by international credit-rating agencies. The greater the percentage, the more the public may be aware of the overall condition of the banking industry as viewed by an independent third party. It takes the value of 1 if the percentage is 100. c) No explicit deposit insurance scheme: this variable takes a value of 1 if there is an explicit deposit insurance scheme and 0 otherwise. d) Bank accounting: this variable takes a value of 1 when the income statement includes accrued or unpaid interest or principal on nonperforming loans and when banks are required to produce consolidated financial statements. e) Off balance sheet items: it takes the value of 1 if off-balance sheet items are disclosed to the public f) Risk management procedures: it takes the value of 1 if banks must disclose risk management procedures to the public g) Subordinated debt: it is equal to 1 if subordinated debt is allowable (required) as a part of regulatory capital. Private monitoring index is the sum of these criteria. Thus, it ranges from 0 to 9, and higher values indicate more private oversight. 27
APPENDIX 3: Capital regulatory index definition by Barth, Caprio, and Levine (2004) It includes two different measures of capital regulatory: a) Overall capital stringency: it measures the extent of regulatory requirements regarding the amount of capital that banks must have relative to specific guidelines. Several guidelines are considered to determine the degree to which the leverage potential for capital is limited. b) Initial capital stringency: it measures the extent to which the source of funds that count as regulatory capital can include assets other than cash or government securities, borrowed funds, and whether the sources of capital are verified by the regulatory or supervisory authorities. Capital regulatory index incorporates the previous two measures of capital stringency. It ranges in value from 0 to 9, with a higher value indicating greater stringency. 28