Is Bank Capital Procyclical? A Cross-Country Analysis

Bikker, Jacob A.; Metzemakers, Paul A. J. Article Is Bank Capital Procyclical? A Cross-Country Analysis Kredit und Kapital Provided in Cooperation with: Duncker & Humblot, Berlin Suggested Citation: Bikker, Jacob A.; Metzemakers, Paul A. J. (2007) : Is Bank Capital Procyclical? A Cross-Country Analysis, Kredit und Kapital, ISSN 1865-5734, Duncker & Humblot, Berlin, Vol. 40, Iss. 2, pp. 225-264, https://doi.org/10.3790/ccm.40.2.225 This Version is available at: https://hdl.handle.net/10419/293569 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Is Bank Capital Procyclical? A Cross-Country Analysis By Jacob A. Bikker and Paul A. J. Metzemakers 1 ,Amsterdam I. Introduction Bank capital plays a pivotal role in bank solvency: the more capital banks have, the more robust their buffers are with which to absorb unexpected losses and, hence, to avoid bankruptcy. It follows naturally that capital is also crucial for accommodating bank lending to firms, which is itself indispensable for healthy macroeconomic development (particularly in bank-based countries): the more capital banks have, the more capacity is available for an expansion in the supply of credit. As is well known, the availability of credit may be threatened when economic conditions grow worse. During a cyclical downturn, the quality of banks' assets generally deteriorates, which increases risk exposure and, hence, economic capital (the total amount of capital needed to cover all risks, as perceived by the institution), exactly at a time when new capital becomes more expensive or, for weaker banks, simply unobtainable. Moreover, loan losses may increase and erode bank capital. As a consequence, banks may be forced to cut back on lending. Particularly in countries where corporate lending is provided mainly by banks, this would further weaken cyclical conditions into a so-called credit crunch, which would in turn exacerbate the downturn. 2 In order to promote bank solvency and to avoid procyclical behaviour by banks, bank supervisors keep an eye on bank capital in relation to Kredit und Kapital 2/2007 Kredit und Kapital, 40. Jahrgang, Heft 2, Seiten 225±264 Abhandlungen 1The authors are affiliated at, respectively, the Strategy Department, Supervisory Policy Division, De Nederlandsche Bank (DNB), and the Economic Policy and Research Division, also at DNB. The views expressed in this article are those of the authors and not necessarily those of DNB. The authors are grateful to Carsten Folkertsma, Aerdt Houben, Jan Kakes, Klaas Knot and other participants of the Financial Stability Research Seminar at DNB for valuable suggestions. 2Strong empirical evidence for the existence of the credit crunch has failed to come up (Sharpe, 1995). See also Berger and Udell (1994), Peek and Rosengren (1995), Wagster (1999) and, for an overview, Bikker (2004), Chapter 7. OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 credit and other risks. One of their tools is to prescribe minimum required capital levels, as has been done since 1988 under the Basel capital Accord, known as Basel I. In 2004, the banking supervisors, gathered in the Basel Committee on Banking Supervision, 3 have agreed on a new capital regime (BCBS (2004)). One of the main objectives underlying the new Basel Agreement is to substantially increase the risk sensitivity of the minimum capital requirements for banks. Earlier draft versions of the Agreement (consultative documents') have prompted a lively debate in both policy circles and the economic literature about the potential procyclical effect such risk-sensitive requirements might have on the economy (e.g. Segaviano and Lowe (2003); Borio et al. (2001); Daníelsson et al. (2001); Carpenter et al. (2001); Turner (2000)). New proposals by the Basel Committee have substantially reduced the possible procyclical effects of the new Agreement and thereby reduced the risks of financial instability. Yet the new capital requirements continue to be more risksensitive than before as, incidentally, they should be in order to promote the financial soundness of banks. Hence, the issue of possible procyclicality continues to exist. In practice, we observe that many banks hold more capital than the required minimum. Many assess their risks independently, for instance, using their own economic capital models. In the trade-off between risk and return, they set an appropriate capital level, depending on their risk aversion. A bank may also prefer to hold more than the required capital as a signal to the market regarding its own soundness in order to be able ± as a very solvent bank ± to raise funds at lower interest rates and for competitive reasons. Banks could also hold buffers as an insurance to avoid costs related to market discipline and supervisory intervention if they approached the regulatory minimum capital ratio (Estrella (2004), Furfine (2000)), or as a cushion to absorb economic recessions, thus limiting the procyclical effect of capital. Finally, banks may respond to regulatory scrutiny by holding higher buffers. Many supervisors require extra buffers in a systematic way or based on individual assessment. 4 All in all, for various reasons, 98% of the banks are above the minimum level, and as many as 86% even hold a quarter more capital than they are re- Kredit und Kapital 2/2007 3The Basel Committee on Banking Supervision consists of senior representatives of banking supervisory authorities and central banks from the (extended) G- 10: Belgium, Canada, France, Germany, Italy, Japan, Luxembourg, the Netherlands, Spain, Sweden, Switzerland, the UK and the US. 4Particularly, supervisors in the US and UK are known for their pressure on banks to hold extra buffers in order to become well-capitalised' banks. 226 Jacob A. Bikker and Paul A. J. Metzemakers OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 quired to. It is not clear in advance whether banks will change their capital buffer behaviour after the changeover to Basel II. There is little doubt that banks will further improve their risk measurement tools and that they may arrive at a more risk sensitive internal risk assessment. Nevertheless, we expect that banks will continue to base their final capital level decisions on their own internal capital targets. It has been argued that a more risk sensitive capital adequacy framework such as Basel II may reduce banks' willingness to take risk. If banks already risk-adjust their total capital, i.e. minimum capital plus buffer capital, more than implied by Basel I, replacing Basel I with Basel II may not affect the capital-to-asset ratio or risk profile of banks' portfolio as much as is feared by some (Lindquist (2004)). For that reason, it is interesting to investigate the cyclical behaviour of banks over the last decade, as this behaviour will probably also be typical for the next decade. Therefore, this article develops a comprehensive model for the possible determinants of bank's own capital target, including business cycle effects. The aim is to detect cyclical patterns in current bank capital behaviour which, if found, might be continued or somewhat amplified under Basel II. A number of recent studies in the economic literature investigate bank capital behaviour. They tend to focus on certain aspects of capitalisation, not always including procyclicality, in a single country only, e.g. Estrella (2004); Lindquist (2004); Ayuso et al. (2004); Rime (2001) and Ediz et al. (1998), respectively on the US, Norway, Spain, Switzerland and the UK. Some more international studies do not focus on procyclicality, but on the impact of Basel I (see the next section). We have opted for a broader approach, applying a comprehensive bank capital target model and using a large data set, which enables us to obtain robust estimates, assuming that similarities in capital behaviour of banks across countries weights heavier than disparities. This approach allows us to compare bank capital behaviour across countries and may reveal world-wide patterns of conduct regarding capital as well as idiosyncratic country-specific deviations. This article is the first that presents such world-wide investigations to bank capital behaviour and procyclicality. For this purpose, we investigate, for each bank, the equity capital level as recorded in its annual reports, that is, including reserves and retained profits, expressed as share of total assets (called equity ratio or leverage ratio) and the capital buffers according to the BIS definitions, that is, its BIS capital as a ratio of so-called risk-weighted assets. The BIS capital ratio is the most interesting one, as it is a risk-adjusted measure of capi- Kredit und Kapital 2/2007 Is Bank Capital Procyclical? A Cross-Country Analysis 227 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 tal and reflects the impact of regulatory requirements more accurately than other capital measures. Unfortunately, the availability of BIS ratio data is rather limited compared to equity capital data. For the equity capital investigations we can employ a large set of over 16,000 bank-year observations from twenty-nine OECD countries over 1990±2001, 5 compared to 7,000 for the BIS ratio analyses. The outline of this article is as follows. Section II reviews minimum capital requirements and actual capital levels against the background of changing supervisory regimes. Section III presents a model for banks' capital levels and constructs proxies for factors that could determine the capital ratio. Section IV elaborates on the data used in the empirical analysis. Section V reports the results of a dynamic multivariate panel regression model for the equity ratio, whereas the next section investigates the BIS capital ratios. Section VII repeats the analyses for various banksize classes in order to assess the equity model's sensitivity to bank sizes. Finally, Section VIII summarises and draws conclusions. II. Banks Capital Buffers and Regulatory Regimes 1. Basel I In 1988, the Basel Committee introduced the first Basel Accord on minimum capital requirements for internationally active banks, in order to promote sound and stable banking systems and a world-wide levelplaying field. At present, over 100 countries have adopted this capital regulatory framework, often also applying it to locally active banks. The BIS or solvency ratio shows a bank's actual own funds (capital) as a percentage of its risk-weighted assets, and must not fall below 8%. The risk-weighted assets relate mainly to the credit risk run by banks, but other risks ± such as market risk ± are also included in the denominator of the BIS ratio. This ratio therefore indicates a bank's capability to absorb losses. However, as not all risks are explicitly taken into account for in the BIS ratio ± take for example operational risk ± banks are required to maintain a capital adequacy ratio of over 8%. The denominator is calculated by multiplying a bank's assets by a weighting coefficient. The greater the (credit) risk, the higher the coefficient. Currently, five coefficients are distinguished: 0%, 10%, 20%, 50% and 100%. The actual own funds forming the numerator of the BIS ratio consist of Kredit und Kapital 2/2007 5The data set does not include Slovakia, which joined the OECD only in 2000. 228 Jacob A. Bikker and Paul A. J. Metzemakers OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 Tier 1, Tier 2 and Tier 3 capital less deductible items. Tier 1 capital, or core capital, tops the list in qualitative terms. It is made up mainly of equity capital, reserves and retained profits, but may, subject to conditions, also include certain innovative forms of capital. At least half of a bank's capital requirement should consist of such core capital, which means that the ratio of Tier 1 capital to risk-weighted assets should be at least 4%. Tier 2 capital is made up of preferred shares and debt certificates with no fixed maturity (upper Tier 2) and of preferred shares with a limited life span and long-term subordinated debts (lower Tier 2). 6 Tier 3, at the bottom of the list in qualitative terms, consists of short-term subordinated debts, and accounts for only a small share of actual own funds. Table 1 presents figures of the BIS capital ratio for our data sample of 7,074 bank-year observations of the BIS ratio, stemming from 1,320 banks. 7 A clear increase of this ratio emerges for the first years from 8.7% on average in 1990 to 10.1% on average in 1994, until in 1995 a tentative equilibrium level has been reached ± which is, incidentally, well above the 8% minimum level. A similar tendency can be observed for the median figures, be it on a higher level. 8 Between 1994 and 2001, the median BIS ratio fluctuated around 12.2%, an ample 50% above the minimum. The persistently higher level of the median reflects skewness of the distribution of the BIS ratio across banks, in the sense that the many small banks tend to maintain higher ratios, whereas the fewer ± more diversified ± large banks maintain lower ratios (see also Table A.5 in the appendix). This is also illustrated by Figure 1, where the frequency distribution of unweighted banks is compared to the frequency distribution of total assets (or size-weighted banks). Kredit und Kapital 2/2007 6Tier 2 also includes, up to certain limits, provisions for general loan loss reserves. This might be a more favourable purpose for retained earnings than equity as, in many countries, such provisions are tax deductible. Bikker and Metzemakers (2005), who investigate bank provisioning behaviour and procyclicality, indeed found a negative relationship between (i) equity and (ii) provisions on the profit and loss account, both taken as shares of total assets. 7The number of observations for the earliest years (1990±1992) and the last year (2001) is much smaller than for the other years. Clearly, the mass of information comes from the central years 1994±2000. Nevertheless, the other years also provide some useful information. 8The tendency of increasing BIS ratios is also reflected in the number of banks that fail to meet the 8% requirement. The share of such weakly capitalised banks falls sharply in the first years, to below 2% in later years. In 1999 this share rises temporarily to above 4%. Is Bank Capital Procyclical? A Cross-Country Analysis 229 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 In the early 1990s only a few banks reported their BIS capital ratio. 9 Using data from national supervisors and the Basel Committee, Jackson et al. (1999) observed that between 1988 and 1992, the transition period, the average capital ratio of the whole sector rose significantly. Apparently, the Accord indeed strongly induced banks to increase their capital reserve. Apart from raising capital, US banks shifted sharply from risky corporate lending to investment in save government securities. There is an extensive amount of literature on this topic due to the fact that the adjustment to Basel I capital levels coincided with a recession in most indus- Kredit und Kapital 2/2007 Table 1 Bank-size Weighted Averages of Annual Capital Ratios (29 OECD Countries) Year BIS capital ratio Equity capital ratio Equity/ BIS Average a Median b Total assets c No. of observations Average a Median b Total assets No. of observations 1990 0.087 0.091 1.2 17 0.042 0.047 1.9 78 0.48 1991 0.091 0.098 2.1 30 0.045 0.058 2.7 152 0.50 1992 0.096 0.100 2.8 64 0.045 0.067 3.4 372 0.47 1993 0.103 0.117 12.4 368 0.044 0.070 14.1 1,306 0.43 1994 0.101 0.125 18.7 759 0.047 0.073 21.7 1,833 0.46 1995 0.105 0.129 18.9 838 0.047 0.076 21.8 1,945 0.44 1996 0.106 0.125 18.7 891 0.048 0.076 22.1 2,030 0.45 1997 0.107 0.123 21.9 906 0.046 0.077 25.3 2,037 0.43 1998 0.115 0.124 23.0 911 0.050 0.075 28.4 2,051 0.44 1999 0.115 0.118 24.7 994 0.053 0.075 29.6 1,993 0.46 2000 0.114 0.117 26.4 914 0.054 0.077 30.3 1,801 0.47 2001 0.120 0.123 9.8 382 0.068 0.086 11.6 428 0.56 All d 0.109 0.122 180.5 7,074 0.050 0.075 212.9 16,026 0.46 a Weighted with total assets; b The median is not weighted. c In thousands of billions of US $; d Here, median is the world wide median: the ratio of the 3,037 th and 8,014 th bank-year observation, respectively. 9This was due to the fact that while industrialised countries adopted riskbased capital standards in 1988, these standards were implemented only gradually, taking full effect as late as 1993. 230 Jacob A. Bikker and Paul A. J. Metzemakers OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 trialised countries. A number of studies made a persuasive case that capital requirements played a role in this switch to less risky assets, supporting the credit crunch hypothesis, 10 but others have provided evidence suggesting that this decline in private lending is better explained by banks' own internal capital targets than by regulatory capital requirements (Hancock and Wilcox (1993); Ediz et al. (1998)). 11 The world-wide (total-assets weighted) average of above 11% indicates that banks choose to maintain capital levels that in almost all cases result in BIS ratios well above the required minimum (see also Figure 1). This outcome underlines that banks may have their own motives for setting capital targets independently from supervisory rules. Banks may be more risk-averse and aim at lower funding costs, they may assess the risk of their portfolio as being higher than the outcome of the BIS risk weighting scheme (De Bondt and Prast (2000)), or they may wish to hold a capital buffer enabling them to exploit unexpected investment opportunities (Berger et al. (1995)). An alternative would be that banks set their capital a certain time-invariant percentage (points) above minimum requirements. Ediz et al. (1998) found for the UK that banks adjust their capital upwards if it comes close to the minimum requirement level or to Kredit und Kapital 2/2007 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 0-2% 2-4% 4-6% 6-8% 8-10% 10-12% 12-14% 14-16% 16-18% 18-20% 20-22% 22-24% 24-26% 26-28% >28% BIS capit al as a percentage of risk weighted asset s percentage of bank observations percentage of total assets Figure 1: Frequency Distribution BIS Capital Ratio (29 OECD Countries, 1990±2001) 10 Hall (1993), Haubrich and Wachtel (1993), Thakor (1996) and Calem and Rob (1999). 11 Other studies investigated whether, within asset categories with equal regulatory risk weights, banks have substituted safer, lower-yielding assets for riskier, higher-yielding investment (Shrieves and Dahl (1992); Haubrich and Wachtel (1993); Jacques and Nigro (1997)). From a theoretical point of view, such substitution can be proven to be sensitive to assumptions about banks objective functions (Rochet (1992a)). Is Bank Capital Procyclical? A Cross-Country Analysis 231 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 the trigger value of the UK supervisors (where the supervisors start drastic actions'). By contrast, Hancock and Wilcox (1993) did not find such adjustments for the US banks. Table 1 also shows the equity capital ratio data, based on 16,026 bankyear observations from 2,536 banks. For two reasons, the (average) ratio is substantially lower than the BIS ratio, in fact, somewhat less than half the BIS ratio. 12 First, equity is equal to the Tier 1 capital, that is, only the highest quality tranche of the buffer capacity. Secondly, the assets in the denominator are not reduced by risk weighting coefficients, ranging from 0% to 100%. The correlation between the BIS and equity capital ratios is 0.65 (with P-value 0.01), significantly different from 1, 13 making it clear that BIS and equity capital ratios often diverge quite strongly. The larger equity ratio sample deviates from the smaller BIS-ratio sample in the sense that the former includes many smaller banks. Hence, these two samples are not fully comparable. 14 The average equity ratio experienced the same rise in the early nineties as the average BIS ratio, be it on a lower level. Apparently, and logically, the rise in the BIS ratio is due mainly to strengthening of the capital structure, rather than to a reduction of the share of risky assets (although such a reduction may have occurred sometimes, see the literature discussion above). The relationship between the equity and BIS ratios has been fairly stable (last column of Table 1). Table 2 presents the BIS and equity ratios for 29 OECD countries. Averaging below 10%, the BIS ratios are lowest in South Korea and Japan, countries known for their banking problems, and in Iceland and Italy. In terms of median values, Germany is also among the countries with the less strongly capitalised banks. The very stable banks in Switzerland and the more risky banks in Turkey, Mexico and Eastern Europe appear to be among the better capitalised ones. The ranking according to equity ratios deviates strongly from that according to BIS ratios. Banks in a number of European countries and in Japan have the lowest equity ratios. Again, Turkey is at the top of the list, followed by Eastern European countries and the US. This indicates serious differences across countries with respect to the capital structure, Kredit und Kapital 2/2007 12 On the other hand, the BIS ratio can be pressed down due to risky off-bal- ance sheet items. 13 By significant (or very significant) we mean, throughout this article, at the 95% (or 99%) level of confidence. 14 This issue is dealt with in the sensitivity to bank-size discussions in Section VII. 232 Jacob A. Bikker and Paul A. J. Metzemakers OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 associate an increase in the loan growth rate with too optimistic expectations about future developments and with diminishing monitoring efforts (Borio et al. (2001); Lowe (2003)). While excessive credit growth may well be related to an increase in risk exposure, this risk will only materialise ± through an increase in problem loans ± with a considerable lag estimated at around three years (Clair (1992); de Lis et al. (2001)). As opposed to total problem loans, these two variables are ex ante risk measures. The signs of their coefficients are expected to be positive as long as banks set their capital in line with the riskiness of their portfolios. However, Rochet (1992a, b) shows that it could be rational for banks with low capital to assets ratios to opt for maximum risk portfolios. If such moral hazard behaviour were widespread, we might find a negative relationship between risk and buffer capital. In addition to Estrella's structural determinants of capital, cyclical variables are used in order to incorporate the effect of economic conditions. Credit risk and losses are negatively correlated with the business cycle. The likelihood of unexpected ± large and infrequent ± credit losses that need to be covered by capital increases during a downturn. Therefore, risk-sensitive behaviour would imply a negative relationship between the level of capital and the business cycle (BC), proxied by the deviation of GDP growth from its country specific average. 24 A positive relation, by contrast, would reflect forward-looking and (also) prudent behaviour, as it would indicate the use of fat years to retain profits for lean years. Another business cycle-related variable is the interest term structure (ITS), the differential between the long-term and short-term interest rate. The opportunity cost of holding capital increases if the interest margin widens, hence we expect a negative relation. An alternative explanation of the interest term structure is that it acts as a proxy of future cyclical fluctuations, which would also imply a negative relationship. Retained earnings are an important source of capital financing and affect the adjustment cost of capital. This justifies the inclusion of returns on assets (ROA), defined as the ratio of after-tax profit and the average of the current and former balance sheets' assets. Its coefficient is expected to be positive. 25 Kredit und Kapital 2/2007 24 An alternative would be GDP growth itself. This variable is probably less precise as the average level of GDP growth may differ across countries. The alternative' empirical results do not deviate much. 25 However, Ediz et al. (1998) assumes a negative relationship, probably assuming that high profits are the consequence of low provisioning and low risk, indi- Is Bank Capital Procyclical? A Cross-Country Analysis 239 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 Bank behaviour may be influenced by a country's legal, regulatory or institutional framework, tax and accounting regime, market or financial structure and business culture, to name a few important country-specific conditions. This was also observed in Table 2. These country-specific characteristics are taken into account by inclusion of dummy variables (d j ) for countries. This brings us to the following equation for capital: (Capital/TA) i,j,t =a 1 +a 2 (Capital/TA) i,j,t±1 +a 3 CROE j,t + (2) a 4 (Customer loans/TA) i,j,t +a 5 Dln Customer loans i,j,t +a 6 BC j,t + a 7 ITS j,t +a 8 ROA i,j,t +S j=1...28 a 8+j d j +e i,j,t Index irepresents individual banks, jthe country in which the bank is located, and trefers to the respective year. Note that all variables are expressed in percentages or are scaled by total assets (TA), which allows comparison across banks of different sizes and across countries. The equation is applied to a world-wide' or OECD sample, the EU and a number of individual countries, as far as ample data are available. A similar model applies to the BIS capital as a ratio of risk weighted assets: we replace the dependent variable Capital/TA and its one-period lagged value in Equation (2) by, respectively, the BIS ratio and its one-period lagged value. Multicollinearity between explanatory variables is not a problem, see Tables A.1 and A.2 in the appendix. The BIS variant of Equation (2) explains the BIS ratio, but can easily be rewritten in terms of a buffer above the minimum BIS requirement as in Ayuso et al. (2004), where the explanatory variable is defined as (BIS- 0.08)/0.08. The latter, after all, is a linear transformation, where Equation (2) is a linear model. The statistical results would, hence, be identical and the coefficients would be a factor 1/0.08 (that is, 12.5) times higher. Similarly, Equation (2) can be rewritten as a capital ratio buffer equation by subtracting the optimal level of the capital ratio K* from its current value K (in terms of Equation (1)). However, we cannot explain such buffers empirically as we cannot observe the optimal level K*. Earlier we observed that banks may hold buffers as (1) a signal to the market regarding its own soundness to raise funds at lower interest rates, (2) an insurance to avoid costs related to market discipline and supervisory intervention if they approached the regulatory minimum capital ratio, (3) a cushion to absorb economic recessions, thus limiting the Kredit und Kapital 2/2007 cating a financial health which allows lower capital. In our view, profit is determined by many factors, provisioning being only one of them. Moreover, higher risk need not harm profits as long as risk is adequately covered by risk premiums. 240 Jacob A. Bikker and Paul A. J. Metzemakers OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 procyclical effect of capital, or (4) because supervisors require extra buffers in a systematic way or based on individual assessment. In general, these buffers help strongly in reducing possible procyclical risks. For that reason, the second pillar of Basel II requires banks to demonstrate that their capital is sufficient to meet the minimum capital requirements during downswings in macroeconomic conditions, forcing them to evaluate risk in a more forward-looking sense. Although we are particularly interested in cyclical determinants of bank capital ratios and possible increased procyclical risks of Basel II, our model is not designed to detect procyclical effects. Instead, our approach is to assess, firstly, whether ample buffers are standard for most banks and, secondly, whether the capital ratio tends to fluctuate with the business cycles. Of course, in our model, various determinants may contribute to cyclical behaviour of the equity ratio, besides the cyclical indicators (GDP growth and interest term structure), in particular loans and returns on assets. The loan portfolio is the major origin of credit risk for which capital is needed. On the other hand, banks might also use loans as an instrument to smooth the equity ratio over time, for instance, by reducing new lending or switching to less risky lending, 26 when the capital ratio decreases. Of course, such policy would increase the risk of procyclicality. This article does not investigate this issue. If, controlled for the effects of other determinants including loans, the ratio would still be positively correlated with the cycle, this could indicate the risk of a lower capital ratio during cyclical downturns, which could again point to increased risk on a credit crunch. If such correlation would be absent or negative, we may expect that buffers indeed function well in cushioning cyclical risks. When banks themselves aim at sufficient buffers and meet the new second pillar requirements, Basel II is not expected to raise procyclical risk substantially. IV. Data and Estimation Approach The analyses are based on pooled cross-section and time series data of individual banks' balance sheet items from 29 OECD countries (listed in Table 2) and country-specific macroeconomic indicators for these countries, over a ten year period from 1992 to 2001. 27 As such, this data set forms a so-called unbalanced panel ± unbalanced as observations are Kredit und Kapital 2/2007 26 The former would make itself felt through the denominator of the capital ratio, whereas the latter would affect the weighting in the BIS ratio. Is Bank Capital Procyclical? A Cross-Country Analysis 241 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 missing due to new entries, mergers or acquisitions, or because of lacking data. The episode 1992 to 2001 covers a full business cycle for all the countries included. Overall, the cycle develops from a trough in the earlier nineties, an economic boom in the mid nineties to a slowdown beginning in 2001. Some countries, such as South Korea, Mexico and Turkey, experienced a severe financial crisis during this period. The macroeconomic data were obtained from the OECD and the IMF (see Appendix 2 of Bikker and Hu, 2002), whereas the balance sheet data were taken from the Bankscope database (Fitch-IBCA). Bank-specific data allow for the investigation of individual banks' capital level characteristics. Moreover, the high number of available observations on banks' capital levels provides a rich source of information. We employed data from commercial banks only, in order to obtain a more homogeneous group of banks. Banks experiencing extreme circumstances are excluded from the sample. Therefore, equity and BIS ratios and loan shares are between 0 and 1, ROA is between ±100% and 100% and loan growth is between ± 80% and 500%. This reduces the sample of the equity ratios by around 4% and that of the BIS ratios by around 3%. The sample selection does not affect the thrust of the estimation results. Similar selections were applied by Cavallo and Majnoni (2002) and Laeven and Majnoni (2003). More details on the data are reported in the appendix. We applied the Generalized Method of Moments (GMM) estimator approach in order to avoid possible biases in the estimates caused by interdependence between the lagged endogenous variable (equity ratio or BIS ratio) and eventual autoregressive terms in the error. This prevents us from making wrong inferences from the t-values, such as regarding significance (Greene (2000)). We estimated Equation (2) but alternatively also considered lags of the explanatory variables. In a number of our regressions, the country's average cost of holding capital (CROE) was lagged more significant than contemporaneous, whereas the contemporaneous CROE was never more significant than lagged. This is plausible as the market's cost of capital is observed with delay, whereas adjustment of equity (or BIS) capital also takes time. Therefore, we applied this lag in all regressions. Lags of other variables did not improve the results and are not shown. The analysis employs a large set of over 16,000 bank-year observations of equity and over 7.000 bank-year observations of the BIS ratio. The Kredit und Kapital 2/2007 27 The basis data cover 12 year (1990±2001), whereas the model uses observations over 10 year (1992±2001) due to lag structures. 242 Jacob A. Bikker and Paul A. J. Metzemakers OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 number of observations per equation in the estimations is much lower, due to the lagged endogenous variables (reduction to around 13,300 and 5,700, respectively) and the second lags of the endogenous variable as instrumental variable used in the GMM procedure (further reduction to 10,500 and 4,300, respectively). Tables A.1±5 in the appendix describe sample statistics. V. Empirical Results for the Equity Capital Ratio We start with the investigation of the equity to total assets ratio, as far more data are available on equity than on the BIS capital ratio. Therefore, the equity model can provide us with the most robust estimation outcomes. The left-hand column of Table 3 presents the estimation results of the dynamic multivariate panel regression Equation (2) applied to the world-wide sample. 28 The disturbances have been tested for serial correlation using the Durbin-Watson (DW) test. 29 The coefficient of the proxy for the cost of capital adjustment is 0.92, reflecting a rather slow adjustment of capital to the target level. This implies that the costs of adjustment are substantial: on average, it takes Kredit und Kapital 2/2007 28 The correlation matrices in Tables A.1±2 in the appendix show that multicollinearity is a problem in neither this regression nor later ones. We tested the model for influences caused by any possible correlation by regressing the lagged dependent and macroeconomic variables first, and then one by one regressing the other variables on the residuals of the preceding regression. The value and significance of all the coefficients remained unchanged. In order to test for stability, we applied a Chow test and re-estimated our model on two sub samples, 1992±1997 and 1998±2001. We observe a slight but significant difference between both subsamples. Granger (1998) explains that common tests loose their validity in the case of very large samples. Similar differences would emerge when the sample has been split along other lines, for example for bank-size classes as in Section VII, or countries as in Tables 3 and 4. A second reason for the significant difference might be that the cyclical effects can be observed less accurately in samples over a few years only, so that we consider the two (short) sub-samples as less suitable. For those reasons we accept the estimation results over the full sample period 1992±2001. 29 The DW test statistic is not applicable to a model with a lagged dependent variable, but the high number of observations made the formula of the appropriate Durbin's h-test statistic intractable. DW test values below 1.60 may well be within the critical limit. This limit is derived from an estimated critical lower limit, D1, which is downward-distorted as the number of explanatory variables (k) increases. The usual DW tables run to k=6 as a maximum with a DW index of 1.57. As this model and subsequent models in this and following tables all contain considerably more than six variables and considerably more observations, we may deduce that the critical D1 is well below 1.57. Is Bank Capital Procyclical? A Cross-Country Analysis 243 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 years before the level of capital is adjusted. 30 Itmayalsoreflectthat bank capital is driven by bank income and default losses rather than by continuous policy adjustments. Apparently, capital reserves adjust more slowly towards their optimal (or model) value than is observed by a similar model for the level of loan loss provisions (0.79 versus 0.92) or for annual additions to provisions (0.41; see Bikker and Metzemakers (2005)). This result supports the capital management view' that provisions might (also) be used to manage the total capital buffer, because provisions can be adjusted more quickly and at lower cost. The coefficient of a country's cost of holding capital (CROE) is negative, as expected (the more expensive capital is, less of it will be held), but not significantly so (at the 95% level of confidence). The risk proxies, customer loan shares and customer loan growth, are both significant, but have counterintuitive negative signs. This has also been observed for Norway by Lindquist (2004). Banks with a relatively risky portfolio do not generally hold more buffer capital. On the contrary, the correlation between equity and loan share is negative (see also Table A.1 in the appendix). Table A.4 shows how banks with high loan shares of 60%±80% maintain the lowest capital ratios, which remains true when weighting by the bank sizes (see the last column), whereas, vice versa, the lowest BIS capital ratios of 8±12% have the highest average loan shares, which, again, remains true when bank are weighted by size (see the average loan share rows). 31 One possible explanation is that the additional credit risk is sufficiently covered by provisions ± or even amply covered as loan loss reserves are usually tax deductible. 32 Larger loan shares and loan growth indeed do significantly increase loan loss provisioning (LLP; Bikker and Metzemakers (2005)). However, inclusion of LLP in Equation (2) does not change the results (not presented), where lower coefficients would have been expected if more risky portfolios were to go hand in hand with higher provisions. If included, the alternative ex post indicator of risk total problem loans' has indeed the expected significantly positive effect on capital (not reported). A disadvantage is that this variable is only available for 40% of the observations. An alternative explanation is that banks may show moral hazard behaviour, that is, that banks with relatively risky portfolios do not in gen- Kredit und Kapital 2/2007 30 Interpreting the model as a weighted average between the old capital level (with weight 0.92) and the optimal model value (with weight 0.08). 31 This holds true for BIS capital, as in Table A.4, as well as equity capital. 32 Note, however, that provisioning covers expected losses whereas capital covers unexpected losses. 244 Jacob A. Bikker and Paul A. J. Metzemakers OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 Kredit und Kapital 2/2007 Table 3 Estimates of the Equity Capital Ratio Model (1992±2001) Variables OECD EU US coefficient t-value a coefficient t-value a coefficient t-value a Intercept 0.0173 **5.3 0.0189 **5.6 0.0072 0.2 Capital, lagged 0.9159 **63.4 0.9225 **46.2 0.8699 **26.2 Country return on equity, lagged ±0.0154 ±1.3 ±0.0348 *±2.0 0.0737 0.3 Customer loan share ±0.0079 **±2.9 ±0.0071 *±2.2 ±0.0025 ±0.4 Customer loan growth ±0.0167 **±8.7 ±0.0175 **±5.7 ±0.0089 **±2.7 Business cycle 0.0124 0.5 0.0700 *2.3 ±0.1360 **±3.0 Interest term structure ±0.0120 ±0.4 ±0.1216 *±2.4 ±0.0310 ±0.2 Return on assets 0.2018 *2.5 0.3982 **5.2 0.1071 0.7 Country dummies Australia ±0.0020 ±1.1 Austria ±0.0036 ±1.6 ±0.0047 ±1.9 Belgium ±0.0070 **±3.0 ±0.0065 *±2.6 Canada ±0.0023 ±0.7 Czech republic ±0.0094 **±3.4 Denmark ±0.0015 ±1.5 ±0.0024 ±1.0 Finland ±0.0090 **±3.5 ±0.0111 **±4.7 France 0.0013 0.6 ± Germany ±0.0015 ±0.7 ±0.0019 ±0.9 Greece 0.0103 *2.5 0.0073 1.5 Hungary ±0.0036 ±1.0 Iceland ±0.0022 ±0.9 Ireland ±0.0073 ±1.8 ±0.0086 ±1.9 Italy ±0.0058 **±3.0 ±0.0076 **±4.1 Japan ±0.0078 **±3.2 Korea, South ±0.0117 **±4.7 Luxembourg ±0.0065 **±3.1 ±0.0053 ±1.9 Mexico 0.0053 0.6 the Netherlands ±0.0038 *±2.1 ±0.0035 ±1.4 Norway ±0.0028 ±1.6 New Zealand ±0.0035 ±1.9 Poland ±0.0016 ±0.3 Portugal ±0.0090 **±4.2 ±0.0101 **±4.3 Spain ±0.0011 ±0.5 ±0.0035 ±1.4 Sweden ±0.0094 *±2.3 ±0.0086 ±1.8 Switzerland 0.0000 0.0 Turkey ±0.0109 **±2.7 UK 0.0006 0.4 ±0.0008 ±0.2 US ± ± No. of observations 10,477 5,681 2,266 Adjusted R 2 0.86 0.86 0.81 Durbin-Watson test statistic 1.64 1.63 1.98 a One and two asterisks indicate 95% and 99% levels of confidence, respectively. Coefficients with one or two asterisks are significant. Is Bank Capital Procyclical? A Cross-Country Analysis 245 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 eral hold more buffer capital. If that were the case, a shift to a more risk sensitive capital regulation regime ± as under Basel II ± would (in principle) force these banks to hold a higher capital level. However, it should be noted that the negative relationship between risk indicators and capital does not necessarily imply that high-risk banks are (too) poorly capitalised relative to the risk in their portfolio. It may rather be due to too much capital in low-risk banks, in line with the substantial capital buffers over the minimum level as observed in Tables 1 and 2. This may reflect strong differences in the way banks evaluate and react to risk, depending on how risk-adverse they are. Another explanation of this negative relationship is that an acceleration of lending ± which would increase both the loan share and the loan growth ± is, initially, financed mainly through additional funding (rather than additional capital), which automatically lowers the capital ratio. This would be in line with the slow adjustment of capital as observed above. Such financing by funding only can typically be expected during booms, when credit risks are assessed to be lower. Such imprudent behaviour would support the theory of Borio et al. (2001) that financial imbalances mount during periods of excessive lending. This conduct would increase the risk of a credit or capital crunch, as the capital buffer is eroded, just before the business cycle bends downwards. The cyclical effects appear to be fairly limited: neither the business cycle indicator BC (defined as the deviation of GDP growth from its country specific average') nor interest term structure has a significant impact. This holds also when these variables are lagged (not reported). One explanation is that the banks' own assessment of credit risk is not sensitive to economic fluctuations. Another is that the other cycle-depen- dent explanatory variables have already picked up this cyclical effect. Finally, return on assets, indicating the ability to retain earnings, has an expected positive effect. The long-run elasticity between capital and ROA is 0.21 so that the capital would increase by at least one fifth, when a bank's profit doubles. 33 Various variables in the equity ratio model may include cyclical patterns: the business cycle indicator, the interest term structure, loan growth and the profit variable, ROA. The multiple regression coefficients provide little or no insight into the net effect of the cycle on the capital Kredit und Kapital 2/2007 33 The long-run elasticities can be calculated given the coefficients of Table 3, the mean values of Table A.1 in the appendix and the lag structure: ±0.2018*0.009/(0.103*(1±0.9159)). 246 Jacob A. Bikker and Paul A. J. Metzemakers OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 level. An alternative approach is to use a simple bilateral comparison between the capital ratio and the business cycle BC, proxied by the deviation of GDP growth from its country specific average', see Figure 2. The classes low', medium' and high' refer to BC values of, respectively, more than 3% below the average GDP growth, around the average GDP growth and GDP growth of more than 3% above the average. 34 Figure 2 shows that fluctuations in BC do not correlate with equity (nor with the BIS ratio). Actually, it is remarkable that GDP growth itself (as an alternative indicator of the business cycle) does ± slightly ± correlate negatively with equity. On average, in periods of GDP growth below 3%, capital is 15% higher (namely 11.7%) than in periods with GDP growth above 3% (where the ratio is 9.9%). 35 This indicates that, hidden behind the various explanatory variables, the capital ratio depends on one of the possible business cycle indicators and suggests that a certain procyclical effect of capital behaviour might exist. Capital appears to depend much less on the business cycle (if at all) than provisioning, which, in periods of GDP growth below 3%, are 60% higher than in periods with GDP growth above 3% (Bikker and Metzemakers (2005)). The BIS capital ratio does not systematically rise or fall with GDP growth, confirming that it does not depend on cyclical fluctuations. All significant country dummies indicate a lower average level of capital compared to the US, the Greek one being the only exception. 36 The higher capital levels of US banks ± after correction for other explanatory Kredit und Kapital 2/2007 0% 2% 4% 6% 8% 10% 12% 14% 16% equit y GDP gro wt h equity BC BIS GDP growt h BIS BC low medium high Figure 2: Relationship Between Business Cycle and Capital Ratios (OECD, 1990±2001) 34 Here, the classes low', medium' and high' refer to GDP growth, respectively, below 0%, between 2 and 4% and above 6%. The two other classes lie in between. 35 The correlation between GDP growth and capital is ±0.08. Is Bank Capital Procyclical? A Cross-Country Analysis 247 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 factors ± could be due to supervisory pressure on adequately capitalised banks to hold 2 percentage points additional capital and become wellcapitalised banks. For example, a bank is well-capitalised if it holds a certain buffer above the adequate levels (which are similar to the minimum requirements of the Basel Accord). This means a bank is well-capi- talised if its leverage or equity ratio is above 6 percent and its BIS-ratio is above 10 percent. Exactly this distinction between adequately and well capitalised may induce US banks to increase their capital ratios above the Basel requirements. Of course, the well-capitalised label may be profitable for a bank as it signals its strength and may lower its cost of funding. In addition, Bikker and Metzemakers (2005) observed that banks outside the US provision more for loan losses. Apparently, US banks hold more capital and provision less than non-US banks. This could be the result of either less risky behaviour on both sides of the balance sheet in the US or more widespread use of general loan loss provisions to increase Tier 2 in Europe. The results of the capital model for the EU ± second column of Table 3 ± have a number of characteristics in common with the world-wide model: the speed of adjustment is similar, the coefficient of returns on assets has a significantly positive sign as expected, the credit risk proxies loan share and loan growth have their imprudent' negative signs, and the coefficients of the EU business cycle and the country-spe- cific return on equity are not significant. There are, however, also differences. For the EU, both the lagged cost of capital proxy, CROE, and the interest term structure variable, reflecting the opportunity cost of holding capital, has a significantly negative sign, in line with expectations. The business cycle indicator, BC, is significantly positive, suggesting prudent forward looking behaviour. Finally, judging by the country dummy coefficients (now in deviation from France instead off the US), also within Europe, differences across countries occur, reflecting diverging accounting and tax rules and other country-specific institutional and economic conditions and behaviour. The right-hand column of Table 3 presents estimates for the US model. Bank capital behaviour in the US differs significantly from that in the OECD and the EU. The cost of adjusting capital ± measured by the speed of adjustment ± is substantial lower in the US than elsewhere. The coef- Kredit und Kapital 2/2007 36 We used data from 29 countries, but only 28 country dummies. The capital ratio of the country with the most bank-year observations ± the US ± is chosen to be reflected by the intercept. 248 Jacob A. Bikker and Paul A. J. Metzemakers OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 could point to eventual moral hazard behaviour. A shift to a more risksensitive capital regulation regime as under Basel II, would (in principle) force these banks to maintain higher ± but still binding ± capital levels, which could imply a behavioural change towards more procyclical capital policies. However, it should be noted that the observed perverse risk effect does not necessarily imply that high-risk banks are (too) poorly capitalised. It may, conversely, be due to low-risk banks holding too much capital, reflecting strong differences between banks in their evaluations of and reactions to risk, or indicating that banks hold buffer capital for other reasons, such as take-over funds. Nevertheless, our analyses point to typically smaller high-risk, low-capital banks, where more risk might exist for shortcoming capital levels during economic downturns when credit risks increase, with potential procyclical effects. The investigations in this article reveal that capital is currently at most moderately cyclical in such a way that procyclical risks might increase slightly. However, for a wide range of reasons banks' own capital targets are generally well above the minimum requirements. Therefore, we expect that under Basel II capital procyclicality will increase only to a limited degree. We can not exclude that some banks will typically set their capital ratio a certain fixed percentage points above the (cyclically dependent) minimum requirements, which would imply cyclicality but not necessary procyclicality. Under the new Agreement, Pillar II requires banks to demonstrate that their capital is sufficient to meet the minimum capital requirements during downswings in macroeconomic conditions, forcing them to evaluate risk in a more forward-looking sense. Strict and adequate stress testing and accompanying capital levels would further reduce the risk of increasing bank capital procyclicality (Peura and Jokivuolle (2004)). This holds in particular for the subset of banks with high loan shares and low buffers above the minimum levels. References Ayuso, J., D. PØrez, J. Saurina (2004): Are capital buffers pro-cyclical? Evidence from Spanish panel data. Journal of Financial Intermediation 13, 249±264. ± BCBS (2004): Basel II: International Convergence of Capital Measurement and Capital Standards: a Revised Framework, Basel Committee Publications No. 107, June (www.bis.org). ± Berger, A. N., G. F. Udell (1994): Do risk-based capital allocate bank credit and cause a credit crunch in the United States?, Journal of Money, Credit and Banking 26, 585±628. ± Berger, A. N., R. J. Herring, G. P. Szegö (1995): The role of capital in financial institutions. Journal of Banking & Finance 19, 393±430. ± Bikker, J. A., H. Hu (2002): Cyclical patterns in profits, provisioning Kredit und Kapital 2/2007 Is Bank Capital Procyclical? A Cross-Country Analysis 255 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 and lending of banks and procyclicality of the new Basel capital requirements, Banca Nazionale del Lavaro Quarterly Review 55, 143±175 (see also: www.dnb.nl). ±Bikker, J. A. (2004): Competition and efficiency in a unified European banking market, Edward Elgar. ± Bikker, J. A., P. A. J. Metzemakers (2005): Bank provisioning behaviour and procyclicality, Journal of International Financial Markets, Institutions & Money. 15, 141±157. ± Bondt, G. J. de, H.M. Prast (2000): Bank capital ratios in the 1990s: cross-country evidence, Banca Nazionale del Lavoro Quarterly Review 53, 71±97. ± Borio, C., C. Furfine, P. Lowe (2001): Procyclicality of the financial system and financial stability: issues and policy options. BIS PapersNo.1,1±57.±Calem, P. S ., R. Rob (1999): The impact of capital-based regulation on bank risk-taking. Journal of Financial Intermediation 8, 317±352. ± Carpenter, S. B., W. Whitesell, E. Zakrajs Ïejk (2001): Capital requirements, business loans, and business cycles: An empirical analysis of potential cyclical effects of the new BIS capital standards, Board of Governors of the Federal Reserve System. ±Catarineu-Rabell, E., P. Jackson, D. P. Tsomocos (2005): Procyclicality and the New Basel Accord ± Banks' choice of loan rating system, Economic Theory 26, 537±557. ± Cavallo, M., G. Majnoni (2002): Do banks provision for bad loans in good times? Empirical evidence and policy implications, in: R. Levich, G. Majnoni, C. Reinhart (eds), Ratings, rating agencies and the global financial system, Kluwer Academic Publishers, Boston, Dordrecht and London. ± Clair,R.T.(1992):Loan growth and loan quality: some preliminary evidence from Texas banks. Federal Reserve Bank of Dallas, Economic Review, 3rd quarter, 9±22. ± Daníelsson, J., P. Embrechts, C. Goodhart, C. Keating, F. Muennich, O. Renault, H.-S. Shin (2001): An academic response to Basel II. Special Paper Series no. 130. ESRC Research Centre. ± Ediz, T., I. Michael, W. Perraudin (1998): The impact of capital requirements on UK bank behaviour. Federal Reserve Bank of New York Economic Policy Review. 4 (3), 15±22. ± Estrella, A. (2004): The cyclical behaviour of optimal bank capital. Journal of Banking & Finance 28, 1469±1498. ± Furfine, C. (2000): Evidence on the response of US banks to changes in capital requirements, BIS Working Papers no. 88, Basel (www.bis.org). ± Granger, C. (1998): Extracting information from mega panels and high frequency data, Statistica Neerlandica 52, 258± 272. ± Greenawalt, M. B., J. F. Sinkey Jr. (1991): Loan-loss experience and risktaking behaviour at large commercial banks. Journal of Financial Services Research 5, 43±59. ± Greene, W. H. (2000): Econometric Analysis. Fourth edition. London: Prentice Hall. ± Hall, B. J. (1993): How has the Basle accord affected bank portfolios? Journal of the Japanese and International Economies 7, 408±440. ±Hancock, D., J. Wilcox (1993): Bank capital and portfolio composition, Bank Structure and Competition, Federal Reserve Bank of Chicago. ± Haubrich, J. G., P. Wachtel (1993): Capital requirements and shifts in commercial bank portfolios, Federal Reserve Bank of Cleveland, Economic Review 29, 2±15. ± Jackson, P. (1999): Capital requirements and bank behaviour: the impact of the Basel Accord, Basel committee on Banking Supervision Working Papers no 1 (www.bis.org). ± Jackson, P. , W. Perraudin, V. Saporta (2002): Regulatory and economic solvency standards for internationally active banks, Journal of Banking & Finance 26, 953± 976. ± Jacques, K., P. Nigro (1997): Risk-based capital, portfolio risk, and bank capital: a simultaneous equations approach, Journal of Economics and Business 49, 533±547. ± Keeley, M. C. (1990): Deposit insurance, risk, and market power in banking. American Economic Review 5, 1183±1200. ± Keeton, W. R. (1999): Does Kredit und Kapital 2/2007 256 Jacob A. Bikker and Paul A. J. Metzemakers OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 faster loan growth lead to higher loan losses? Federal Reserve of Kansas City, Economic Review, 2nd quarter. 57±75. ± Laeven, L., G. Majnoni (2003): Loan loss provisioning and economic slowdowns: too much, too late? Journal of Financial Intermediation 12, 178±197. ± Lindquist, K.G.(2004):Banks'buffercapital.Howimportant is risk? Journal of International Money and Finance 23, 493±513. ± Lis, F. S. de, J. Martinez PagØs, J. Saurina (2001): Credit growth, problem loans and credit risk provisioning in Spain. BIS Papers no 1, 331±353. ± Lowe, P. (2003): Credit risk measurement and procyclicality, in: T. G. Kuppens, H. M. Prast, A. A. T. Wesseling (eds), Supervision on a Cross-road, Edward Elgar, Cheltenham, UK, and Northampton MA, USA, 136±163. ± Lowe, Ph., M. A. Segoviano (2002): Internal ratings, the business cycle and capital requirements: some evidence from an emerging market economy. Bank for International Settlements, Basel, mimeo. ± Peek, J., E. Rosengren (1995): The capital crunch: neither a borrower nor a lender, Journal of Money, Credit, and Banking 27, 623±638. ± Peura, S., E. Jokivuolle (2004): Simulation based stress test of banks' regulatory capital adequacy, Journal of Banking & Finance 28, 1801±1824. ± Rime, B. (2001): Capital requirements and bank behaviour: empirical evidence for Switzerland, Journal of Banking & Finance 25, 789±805. ± Rochet, J. C. (1992a): Capital Requirements and the behaviour of commercial banks, European Economic Review 36, 1137±1170. ± Rochet, J. C. (1992b): Towards a Theory of Optimal Banking Regulation, Cahiers Economiques et Monetaires no. 40. Banque de France, 275±284. ± Segaviano, M. A., P. Lowe (2002): Internal ratings, the business cycle and capital requirements: some evidence from an emerging market economy, BIS Working Paper no. 117, Basel. ± Salas, V., J. Saurina (2003): Deregulation, market power and risk taking in Spanish banks. European Economic Review 47, 1061±1075. ± Sharpe, S. A. (1995): Bank capitalisation, regulation, and the credit crunch: a critical review of the research findings, Finance and Economic Discussion Series 95/20, Board of Governors of the Federal Reserve Board, 1±23. ± Shrieves, R. E., D. Dahl (1992): The relationship between risk and capital in commercial banks, Journal of Banking & Finance 16, 439±457. ± Thakor, A. V. (1996): Capital requirements, monetary policy, and aggregate bank lending: theory and empirical evidence, Journal of Finance 51, 279±324. ± Theil, H. (1971): Principles of Econometrics, John Wiley & Sons, Inc., New York, London. ± Turner, P. (2000): Procyclicality of regulatory ratios? Center for Economic Policy Analysis. Working Paper no. 13. ± Wagster, J. D. (1999): The Basel accord of 1988 and the international credit crunch of 1982±1992. Journal of Financial Services Research 15, 123±143. Kredit und Kapital 2/2007 Is Bank Capital Procyclical? A Cross-Country Analysis 257 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 Kredit und Kapital 2/2007 APPENDIX: Data Statistics Table A.1 Data Statistics of Determinants of the Equity Ratio Model (all OECD Countries) Equity capital ratio Idem, lagged Country return on equity, lagged Cust. loan share Cust. loan growth Business cycle Interest term structure Return on assets Mean 0.103 0.102 0.097 0.518 0.126 0.005 0.011 0.009 Median 0.076 0.075 0.091 0.545 0.057 0.007 0.014 0.006 Standard deviation 0.104 0.104 0.069 0.251 0.462 0.015 0.021 0.019 No. of observations 10,477 Correlation matrix Equity capital ratio 1.00 Idem, lagged 0.92 1.00 Country return on equity, lagged ±0.02 ±0.01 1.00 Customer loan share ±0.18 ±0.18 ±0.03 1.00 Customer loan growth 0.00 0.08 0.07 0.03 1.00 Business cycle 0.00 0.01 0.19 0.02 0.07 1.00 Interest term structure (%) ±0.02 ±0.03 ±0.23 0.02 ±0.11 ±0.18 1.00 Return on assets 0.27 0.26 0.20 ±0.02 0.08 0.06 ±0.12 1.00 Explanation: Data statistics refer to estimation sample of first column of Table 3. 258 Jacob A. Bikker and Paul A. J. Metzemakers OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 Kredit und Kapital 2/2007 Table A.2 Data Statistics of Determinants of the BIS Capital Ratio Model (OECD Countries) BIS capital ratio Idem, lagged Country return on equity, lagged Loan share Loan growth Business cycle Interest term structure Return on assets Mean 0.145 0.146 0.114 0.581 0.154 0.006 0.012 0.010 Median 0.122 0.123 0.132 0.610 0.082 0.008 0.015 0.010 Standard deviation 0.072 0.077 0.065 0.190 0.420 0.014 0.013 0.012 No. of observations 4,287 Correlation matrix BIS capital ratio 1.00 Idem, lagged 0.86 1.00 Country return on equity, lagged 0.07 0.08 1.00 Customer loan share ±0.33 ±0.33 0.13 1.00 Customer loan growth ±0.05 0.05 0.07 0.07 1.00 Business cycle 0.03 0.06 0.17 0.01 0.06 1.00 Interest term structure (%) 0.01 0.00 ±0.18 0.04 ±0.07 ±0.13 1.00 Return on assets 0.19 0.19 0.27 0.11 0.11 0.08 ±0.14 1.00 Explanation: Data statistics refer to estimation sample of first column of Table 4. Is Bank Capital Procyclical? A Cross-Country Analysis 259 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 Kredit und Kapital 2/2007 Table A.3 Median of Bank-specific and Economic Model Variables per Country AUS AUT BEL CAN CZE DEN FIN Equity capital ratio 0.061 0.062 0.050 0.075 0.066 0.115 0.049 BIS capital ratio 0.109 0.105 0.120 0.116 0.126 0.136 0.125 Country return on equity, lagged 0.162 0.080 0.105 0.088 0.088 0.135 0.067 Customer loan share 0.800 0.492 0.296 0.734 0.411 0.598 0.487 Customer loan growth 0.045 ±0.004 0.013 0.040 0.164 0.017 0.044 Business cycle 0.008 0.002 0.002 0.015 ±0.010 0.005 0.022 Interest term structure (%) 0.015 0.013 0.023 0.014 0.007 0.016 0.017 Return on assets 0.007 0.004 0.004 0.004 0.005 0.011 0.004 Number of observations 175 187 212 224 96 307 39 FRA GER GRE HUN IRE ITA JPN Equity capital ratio 0.067 0.067 0.089 0.097 0.060 0.074 0.041 BIS capital ratio 0.112 0.099 0.125 0.153 0.122 0.116 0.102 Country return on equity, lagged 0.024 0.062 0.145 0.229 0.160 0.034 0.017 Customer loan share 0.507 0.466 0.420 0.446 0.583 0.511 0.694 Customer loan growth ±0.005 0.000 0.161 0.122 0.140 0.066 ±0.122 Business cycle ±0.001 ±0.004 0.011 0.021 0.028 0.001 0.012 Interest term structure (%) 0.016 0.016 ±0.005 ±0.025 0.011 0.009 0.014 Return on assets 0.004 0.002 0.012 0.014 0.006 0.005 0.001 Number of observations 1,258 1,112 59 118 87 498 280 260 Jacob A. Bikker and Paul A. J. Metzemakers OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 Kredit und Kapital 2/2007 KOR LUX MEX NETH NOR NZEA POL Equity capital ratio 0.047 0.036 0.124 0.070 0.065 0.050 0.110 BIS capital ratio 0.098 0.119 0.137 0.134 0.108 0.107 0.142 Country return on equity, lagged 0.055 0.036 0.090 0.117 0.184 0.203 0.208 Customer loan share 0.507 0.151 0.584 0.482 0.847 0.811 0.487 Customer loan growth 0.100 ±0.012 0.170 0.109 0.073 0.100 0.283 Business cycle 0.008 0.017 0.016 0.006 0.011 ±0.007 0.003 Interest term structure (%) ±0.017 0.018 0.017 0.025 ±0.003 0.005 ±0.021 Return on assets 0.002 0.004 0.002 0.005 0.010 0.011 0.014 Number of observations 120 627 100 210 58 39 163 POR SPA SWE SWI TUR UK USA Equity capital ratio 0.062 0.079 0.039 0.119 0.084 0.088 0.084 BIS capital ratio 0.112 0.109 0.122 0.144 0.190 0.160 0.123 Country return on equity, lagged 0.066 0.087 0.153 0.040 0.242 0.194 0.133 Customer loan share 0.492 0.515 0.582 0.546 0.394 0.453 0.648 Customer loan growth 0.095 0.066 0.042 0.014 0.242 0.090 0.099 Business cycle 0.010 0.014 0.018 0.031 0.038 0.009 0.008 Interest term structure (%) 0.012 0.008 0.019 0.014 ±0.132 0.002 0.015 Return on assets 0.005 0.007 0.004 0.007 0.022 0.009 0.012 Number of observations 171 383 42 1,008 128 489 2,266 Explanation: Data statistics refer to estimation sample of first column of Table 3 (BIS ratio: first column of Table 4). Iceland has been omitted as their number of bank-year observation is too low. Is Bank Capital Procyclical? A Cross-Country Analysis 261 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 Kredit und Kapital 2/2007 Table A.4 Distribution of Number of Banks and Total Assets over BIS Capital Ratio and Customer Loan-share Classes Loan share BIS ratio Avgs BIS ratio a 8%±10% 10%±12% 12%±14% >14% Total Number of banks 0%± 40% 1.8 3.0 2.5 11.3 18.5 20.4 40%± 60% 4.1 8.8 6.1 12.1 31.1 14.5 60%± 80% 5.6 17.8 8.6 8.8 40.8 12.6 80%±100% 1.5 4.0 1.8 2.4 9.6 13.1 Total 13.0 33.5 18.9 34.6 100.0 14.7 Avgs loan share 60.0 63.2 59.6 48.4 57.1 Total assets 0%± 40% 2.4 7.6 3.5 3.4 16.9 12.5 40%± 60% 9.9 18.6 5.6 2.0 36.2 11.0 60%± 80% 22.1 16.3 5.2 0.8 44.4 10.3 80%±100% 0.3 1.2 0.5 0.6 2.5 13.3 Total 34.8 43.6 14.8 6.8 100.0 10.9 Avgs loan share 61.3 55.2 52.0 41.5 55.7 Explanation: The table is based on data of banks from all OECD countries over 1990±2001. a Respectively, unweighted (upper part) and weighted with total assets (lower part). 262 Jacob A. Bikker and Paul A. J. Metzemakers OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 Kredit und Kapital 2/2007 Table A.5 Distribution of Number of Banks and Total Assets over BIS Capital Ratio and Bank-size Classes Bank size BIS ratio Avgs BIS ratio a 8%±10% 10%±12% 12%±14% >14% Total Number of banks Small b 5.1 17.8 10.4 27.5 60.9 16.4 Medium 6.4 13.3 4.9 4.2 28.8 11.3 Large 2.1 5.6 1.9 0.7 10.3 11.4 Total 13.6 36.7 17.2 32.5 100.0 14.3 Total assets Small 0.5 1.5 0.7 1.3 3.9 14.6 Medium 4.6 10.3 3.0 2.1 20.1 11.9 Large 10.5 39.1 20.6 5.8 76.0 11.5 Total 15.6 50.9 24.2 9.3 100.0 11.6 Explanation: The table is based on 1999 data of banks from all OECD countries. a Respectively, unweighted (upper part) and weighted with total assets (lower part). b Large banks have in 1999 a balance sheet total of above US $ 44 billion and small banks have a balance-sheet total of below US $ 3.1 billion. These borders are twice or triple times those of the equity ratio regressions in Section VII, as many small banks report capital ratios but do not report BIS ratios. Is Bank Capital Procyclical? A Cross-Country Analysis 263 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28 Summary Is Bank Capital Procyclical? A Cross-Country Analysis This article investigates the determinants of commercial banks' own internal capital targets and potential sensitivity of these levels to the business cycle. Worldwide results make clear that banks' own risk is only slightly dependent on the business cycle. Banks tend to hold substantial capital buffers on top of minimum requirements, reflecting that they hold capital for other reasons than strictly meeting the capital requirements. These results suggest that actual capital levels may not become substantially more procyclical under the new risk-sensitive Basel II regime. However, a number of banks, especially smaller ones, combine a relatively risky portfolio with limited buffer capital. A more risk-sensitive capital regulation regime could force these banks to obtain higher capital levels, which would make them more procyclical. (JEL E32, G21, G28, G31) Zusammenfassung Wirkt Bankenkapital prozyklisch? Eine länderübergreifende Untersuchung In diesem Artikel werden die Bestimmungsfaktoren für die internen Eigenkapitalziele von Geschäftsbanken sowie die potenzielle Sensitivität der unterschiedlichen Kapitalhöhen im Konjunkturzyklus untersucht. Weltweit erhobene Daten zeigen klar, dass das Eigenrisiko der Banken nur geringfügig vom Konjunkturzyklus beeinflusst wird. Die Banken neigen dazu, über das für die Unterlegung mit Eigenmitteln erforderliche Mindestkapital hinaus weiteres Pufferkapital in substanzieller Höhe zu halten, woraus hervorgeht, dass sie Kapital auch für Zwecke halten, die nicht ausschlieûlich der strengen Einhaltung der Bestimmungen über die Unterlegung mit Eigenmitteln dienen. Diese Daten legen die Vermutung nahe, dass die derzeitigen Kapitalhöhen gemäû dem neuen risikosensitiven Basel-II- Regime nicht wesentlich stärker prozyklisch wirken dürfen. Jedoch kombiniert eine Reihe von insbesondere kleineren Banken relativ stark risikobehaftete Portfolios mit Pufferkapital in begrenzter Höhe. Ein stärker risikosensitives Regime für die Unterlegung von Krediten mit Eigenmitteln könnte diese Banken dazu zwingen, höhere Eigenmittel vorzusehen, wodurch deren Wirkung stärker prozyklisch sein würde. Kredit und Kapital 2/2007 264 Jacob A. Bikker and Paul A. J. Metzemakers OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.40.2.225 | Generated on 2023-01-16 13:24:28