You are on page 1of 15

Journal of Banking & Finance 37 (2013) 3373–3387

Contents lists available at SciVerse ScienceDirect

Journal of Banking & Finance
journal homepage: www.elsevier.com/locate/jbf

Banks’ capital buffer, risk and performance in the Canadian banking
system: Impact of business cycles and regulatory changes
Alaa Guidara a, Van Son Lai a,⇑, Issouf Soumaré a, Fulbert Tchana Tchana b
a
Department of Finance, Insurance and Real Estate, Faculty of Business Administration, Laval University, 2325 rue de la Terrasse, Quebec, Canada G1V 0A6
b
Ministry of Finance, 12 rue Saint-Louis, Quebec, Canada G1R 5L3

a r t i c l e i n f o a b s t r a c t

Article history: Using quarterly financial statements and stock market data from 1982 to 2010 for the six largest Cana-
Received 31 January 2013 dian chartered banks, this paper documents positive co-movement between Canadian banks’ capital buf-
Accepted 6 May 2013 fer and business cycles. The adoption of Basel Accords and the balance sheet leverage cap imposed by
Available online 15 May 2013
Canadian banking regulations did not change this cyclical behavior of Canadian bank capital. We find
Canadian banks to be well-capitalized and that they hold a larger capital buffer in expansion than in
JEL classification: recession, which may explain how they weathered the recent subprime financial crisis so well. This evi-
G21
dence that Canadian banks ride the business and regulatory periods underscores the appropriateness of a
G28
both micro- and a macro-prudential ‘‘through-the-cycle’’ approach to capital adequacy as advocated in
Keywords: the proposed Basel III framework to strengthen the resilience of the banking sector.
Capital buffer Ó 2013 Elsevier B.V. All rights reserved.
Risk
Performance
Basel accords
Regulation
Business cycles
Canadian banks

1. Introduction requirement of the Office of the Superintendent of Financial Insti-
tutions (OSFI). Cyclicality of bank capital is defined as the co-
The 2007 subprime turmoil underscores the imperative for a movement between business cycles and bank capital. Positive co-
sound micro- and macro-prudential framework for banking regula- movement implies counter-cyclicality and negative co-movement
tion and supervision to build up resilience against severe crises and denotes procyclicality.2 Therefore, to have counter-cyclicality be-
to ensure the stability of the entire financial system.1 During this tween bank capital buffers and the business cycle, capital has to be
crisis, Canada’s banking system performed much better than other accumulated in booms and lower in troughs.3 Second, we analyze
industrialized countries. Even as high-profile banks in Europe, the the impact of capital buffers on banks’ risk and performance, control-
United States and elsewhere collapsed, were bailed out, or under- ling for business cycles as well as for capital regulatory environ-
went imposed take-overs—Fortis, Citigroup, UBS and the Royal Bank ments, namely in the period preceding the Basel Accords, during
of Scotland are a few examples—not one Canadian bank failed or was Basel I, and during amendments to the Basel I and Basel II regimes.
openly bailed out.
In this paper, we examine the relationship between bank capital
2
buffers and business cycles in Canada’s banking sector. We first See for instance, Illing and Paulin (2004).
3
Noteworthy empirical evidence on the dependence of capital buffer on business
examine the cyclicality of Canadian banks’ capital buffer with re-
cycles, albeit still limited and conflicting, is as follows. Lindquist (2004), Jokipii and
spect to business cycles, where the buffer (excess capital) is the Milne (2008, 2011) and Stolz and Wedow (2011) find a negative co-movement of
size of the capital cushion that exceeds the regulatory capital capital buffers of Western European banks and the business cycle. Shim (2013) also
documents a negative relationship between the business cycle and US banks’ capital
buffers, implying that banks may increase capital buffers by shrinking their risk-
⇑ Corresponding author. Tel.: +1 418 656 2131x3943; fax: +1 418 656 2624. weighted assets during business downturns. Further, Jokipii and Milne (2008) find
E-mail addresses: alaa.guidara@fsa.ulaval.ca (A. Guidara), vanson.lai@fas.ulaval. that in recessions, large banks, commercial and savings banks increase their capital
ca (V.S. Lai), issouf.soumare@fsa.ulaval.ca (I. Soumaré), fulbert.tchanatchana@ buffer, while small banks and co-operative banks as well as those in accession
finances.gouv.qc.ca (F.T. Tchana). countries decrease their capital buffer, suggesting that the cyclical behavior of bank
1
Micro-actions pertain to management actions at the bank level. Macro-actions capital buffers varies according to the size, the type of bank, the country financial
refer to monetary and other policies at the country level or higher. infrastructure and regulatory environment.

0378-4266/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jbankfin.2013.05.012

vide one possible explanation for how Canadian banks weathered We use the following system of simultaneous equations: the recent financial crisis better than banks in other countries. Since then. following the 1996 amendment to the Basel I. RISKj. the Bank Act’s review period We can then draw two main policy implications from the Cana.t ¼ f 3ðCR3t . Since the negative co-movement between capital buffer where the dependent variables are as follows: DBUFj. the Bank both risk-based and non-risk-based capital requirements may help Act required banks to account for market risk when computing mitigate the well-documented procyclicality associated with cur. Canada enforced the Basel II requirements starting in November versely. cost frontiers) endogenously.09%) and the leverage capital buffer (0. although it is  DREGt . dian experience. (2007).  DREGt . In exploring ð1Þ the role played by the Basel regulations in this relationship. Jacques and Nigro (1997). We conclude in Section 6. Third. Canadian banks have accounted Q are used as performance measures. In Section 5. PERFj.t . tutions (OSFI). we find that this positive co-movement is still present after the 1996 DRISKj.. and higher liquidity ratios (e.t1 . This finding is similar to below. Regulatory background and the leverage constraints imposed by Canadian regulators. be reassessed and updated each decade (Calmès.t . 4 5 Other reasons include conservative mortgage practices. Our results support the view that Basel 2.. tions with significant trading portfolios are required to maintain capital to cover Ratnovski and Huang. 2004). OUTGAPt . of risk of bank j at time t. dealers. we carry out robust- Our work departs from the literature on capital buffers in sev.t . Basel II introduced operational risk into the RWA calculation ments should be higher during booming economic periods and proposed the internal ratings-based approach for credit risk. we describe the data How do induced changes in bank capital buffers affect the perfor. BUFRj.g. banks to operate. we depart from it. We also find a negative but not bank j at time t.t . When equity returns are used to measure performance. TERMt . / Journal of Banking & Finance 37 (2013) 3373–3387 Our research questions are as follows: (1) Do Canadian banks’ capital periods would be welcome since this may provide more room for buffers run counter to business cycles? (2) Are Canadian banks’ cap. risk and performance (2001) and others have used systems of two simultaneous equa- simultaneously. DREGt . To contrast. OUTGAPt banks’ capital buffer and business cycles (countercyclical effects):  DREGt Þ. Northcott et al. This amendment started to be enforced in 1998.4 We also document positive co-movement between Canadian DBUFj. 2009. unlike some previous research. more pronounced over the 1988–1997 Basel I period. have to some extent succeeded at better aligning Canadian forced by Canada’s Office of the Superintendent of Financial Insti- banks’ risk-taking with their capital base. In addition. Stolz and Wedow (2011)) and on US banks (e.t . another amendment to the Bank Act allowed banks to Q. CVj. in business cycles and multiple regulatory changes. However. DRISKj. Shim (2013)) find ð3Þ a negative co-movement between business cycles and banks’ cap- ital buffer. In Section 2. CREDITj. Kwan and Eisenbeis (1997) and Altunbas et al.1.t1 .t1 . DBUFj. most studies on Euro. our study period covers at least three regulatory environments. This law was passed in 1871 and was supposed to Finally. First.t . OUTGAPt . this is the first paper to comprehensively address these is- study banks’ capital. was shortened from 10 to 5 years (Calmès. DREGt .t  DREGt Þ.t . DREGt .t is the varia- and business cycles can exacerbate the procyclical impact of Basel tion of the capital buffer of bank j at time t..49%). ital buffers sensitive to changes in capital regulations? (3) How sen.g.g. market risks. The 1987 mance of Canadian banks depends on how performance is mea. In Section 3. DBUFj. In Section 4. Basel regulations introduced credit risk-based there is no effect. these studies underline the need for capital provision. it uses an extensive database of quarterly data over a relatively long period (1982–2010) to study Canada’s banking 2. their RWA. because this is when banks can accumulate more capital. 2011). 2004). This may ð2Þ be one explanation for the resilience of the Canadian banking sec- tor to the recent financial crisis. we give a brief overview that of Lindquist (2004). exceed the lows the latter approach. Rime the relationship between capital buffers. But before describing the variables. we find that the impact of capital buffer on the perfor. icant and positive impact on ROA and a negative impact on Tobin’s In 1992. eral ways. more capital is being accumulated during booms. Note that while our specification fol- We find that Canadian banks are well-capitalized. first by focusing on minimum requirements for both the regulatory capital buffer capital buffers instead of capital ratios and second by superimpos- (5. DBUFj. DRISKj. Jokipii and Milne (2008.t the variation regulation. DBUFj. buy trust companies. TERMt . These findings pro- ing the effect of business cycles under banking regulation changes.t . DPERFj. In 1988. capital require. DRISKj. who found support for the hypothesis that of the regulatory background in Canada. if returns on assets (ROA) or Tobin’s capital requirements.t ¼ f2 ðVTSXt .t . Second. To our knowl- contrast. we study Shrieves and Dahl (1992). deposit-taking institu- market wholesale funding. Empirical framework sector. developing a system of three simultaneous equa- tions to study the relationship between banks’ risk and their capi- tions that link capital buffer. for this risk when calculating their risk-weighted assets (RWA). we discuss and mance of Canadian banks? interpret the empirical results. rigorous and disciplined implementation of In 1997. DPERFj. capital buffers may be considered as insurance against failure to meet capital requirements. OUTGAPt . prin- cipally the Office of the Superintendent of Financial Institutions Canada’s banking sector is regulated by the Bank Act and is en- (OSFI). OUTGAPt pean banking institutions (e. ness checks. 2009).t DDREGt Þ.5 In rent Basel risk-based capital charges.t the variation of performance of ing during good economic times. statistically significant relationship between variations in banks’ These variables and the other explanatory variables are defined capital buffer and banks’ risk exposure. a reduction in capital requirements during recessionary 2007.. Secondly. 2004. formulated systems of three simultaneous equations to edge. and present the descriptive statistics. DPERFj. BUFj. Guidara et al. The rest of this paper is structured as follows.’’ .3374 A. risk and efficiency (derived from stochastic sues relating to capital buffers in the Canadian context. Con. non-reliance on money An OSFI report states that ‘‘Beginning January 1st 1998.t . risk and performance within several tal. OUTGAPt amendment to the Basel I Accord adopted in 1998. we sitive are Canadian banks’ risk to changes in their capital buffer? (4) discuss our empirical framework. First. DPERFj.t ¼ f1 ðSIZEj. capital buffers have a signif. amendment to the Bank Act allowed banks to acquire investment sured.t . SIZEj.

however.t  DREGt is the cross-product of DBUFj. – OUTGAPt is a business cycle indicator. and denote it as BUFR.t represents the log of total assets of bank j at time t and between CAP and the minimum regulatory capital requirement controls for the size effect (Jacques and Nigro.jUx. these measures are thus less precise than BUFL. Jokipii. the error term. 2002. It has been calculated as the standard deviation of j.jUI. where V is the implicit total asset value (the first DREG2 controls for the 1997 amendment to the Bank Act and unknown). lue (V) and its implicit volatility (rV) are obtained by solving a system – DREGt are dummy variables to control for the stages of Basel of equations based on shareholders’ equity defined as a call option: regulations. We expect idiosyncratic risk (IRISK) and the implicit volatility of assets this variable to positively impact performance. and captures the interaction have also used this constant in their study of bank moral hazard and between variations in capital buffers and the regulatory regimes. regulations and business cycles. We calculate TRISK using the standard deviation of daily – CVj. with x ¼ ½LnðV=qBÞ þ ðr2V T=2Þ=rV T elsewhere to reflect Based I regulations before the amendment. 1) in mean of conditional volatility on the residual from a mul.t represents the ter. we dropped these additional GJR.7 We add an additional factor value of equity. The index includes the six Canadian chartered banks in our interest rate risk premium computed as the difference between sample.) is the standard cumulative normal distribution ulatory regime dummy DREGt. BUFR and BUFE. 9 be best.t is the charter value used to control for banks’ incentives for equity returns over the quarter. captures shocks on the obtained using the approach of Ronn and Verma (1986). 1990). to keep the model simple. BUFL is our main capital buffer measure because it is easy to – GDPGt is the quarterly growth rate of real GDP. we added variables such as capital issuance. DREG1 takes a value of 1 over 1988–1997 and zero pffiffiffi pffiffiffi K ¼ VNðxÞ  qBNðx  rV T Þ.t is the ratio of total loans over total assets and is used change rate of the Canadian dollar to the US dollar (the US dollar to control for the impact of lending activities on a bank’s capital is the most commonly used foreign currency in Canada) and ej. Our use alternative performance metrics: (i) the return on assets (ROA) first and main measure of the capital ratio is the leveraged capital obtained as the ratio of net income over total assets. Rime. and BVE is the book and Théoret (2010). The 1997 amendment to the Bank Act came at bank’s total debt.t + bx. We calculate IRISK using a GARCH risk-taking (e. which measures a We use the following explanatory variables8: bank’s capital-to-risk-weighted assets (RWA) ratio.. where Rj.6 The economic capital buffer BUFE is obtained as the dif. B is the book value of the Basel II effects. A. dividend which supposes a credit rating of at least AA+ for each bank of the sample.2. among others. 0. we use the banks’ mean daily stock market returns (RET) over the last calendar quarter. Rm. follows: tifactor market model over the last quarter of daily observations. We expect this variable to negatively We also compute a third capital ratio.jUI.g.t and the regula- American and Canadian banks. among others).t. MVE is the market value of equity. Guidara et al.g. are difficult – CR3t is the income concentration ratio at time t computed as the to obtain due to the lack of complete and comprehensive informa.t = b0. (ARISK). / Journal of Banking & Finance 37 (2013) 3373–3387 3375 2. K is the market value of equity. reinvested earnings and external financing to the original model asset values from contingent claim analysis. rV is the unobserved bank asset return volatility (the second un. proxy industrial concentration and competition in the banking We use three risk measures: total equity risk (TRISK). Our second capital ratio variable is CAP. market risk. the difference between the yield on long-term Canadian The risk measure ARISK is the implicit volatility of asset returns (rV) government bonds and the T-bill yield. market sector (e. DREG2 takes a turns.jRm.. rK is the standard deviation of the bank’s equity re. It is the inverse of the balance sheet leverage ratio and Q (QTOB) computed as the market value of equity divided by its is obtained as the ratio of shareholders’ book equity over total as.jRm. . We derive policy parameters. of real gross domestic product (GDP) obtained by applying the ference between the bank’s actual capital ratio and its economic Hodrick–Prescott filter. ratio of the total net income of the three largest banks divided tion and data to compute the risk-weighted assets and VaR of by the total net income of the sector. (1997) and Calmès of assets.97% confidence level. known). risk and performance measures As a measure of performance. – CREDITj.t + bI. and Ln is the logarithmic operator. (CAPE). value of 1 from 1998 to 2010 and zero elsewhere. 2006).. ship between this market risk and our six banks’ risk measures.t + bm. the economic capital ratio impact the variation in the capital buffer and performance. and (ii) Tobin’s ratio (CAPL).t + bI. T is the maturity of the debt (we – OUTGAPt  DREGt is the cross-product of OUTGAPt and the reg- assume 1 year). using the value at risk (VaR) based on the bank’s asset dis. and found that results did not differ significantly from the results obtained from the 7 We run several other conditional volatility specifications such as EGARCH and basic model. Evaluating the results with Akaike and Schwarz’s criteria. a proxy for Canadian et al.t + bm. – We also use bank dummies to control for bank-specific effects.. It is the cyclical component tribution. Keeley. UI. Flannery et al. Beck et al. respectively. and captures the interaction function. – CV = Ln((BVA + MVE  BVE)/BVE). Bikker and Haaf. We use the cyclical output gap instead capital ratio. term structure of interest rates.3.t = b0.t is the market premium. where BVA is the book value This is similar to Song (1994). It is calculated as (1. Gueyie and Lai (2003) tory regime dummy DREGt. q is a regulatory parameter.t is buffer. as in Ronn and Verma (1986).t + ej. The buffer with this capital ratio measure is denoted as BUFL and is measured by CAPL minus the inverse of the balance sheet leverage ratio cap fixed by Cana. 1997. of real GDP because it removes trends from time series variables. – TERMt. (1989) for – DBUFj. 8 To control for the active adjustment of capital buffers in response to banking 6 We compute VaR using the asset distribution at the 99. as well as many other firms. sets. the introduction of deposit insurance in Canada.97 as in Ronn and Verma (1986) and Giammarino et al.t is the equity return daily returns of the S&P/TSX Composite Index9 over the last quar- of bank j at time t. Capital buffers. The parameter q equals between business cycles and the regulatory regimes. We also We use three capital ratio measures to compute the buffer.t is the exchange rate premium computed as one minus the ex. 2009. Explanatory variables dian banking regulations. we find GARCH-M to variables. Total asset va. 2. for exchange rate risk to the market multifactor model used by Chen – VTSXt is the volatility of the market index. This variable is used to Canadian banks. and rK = rV V N(x)/K. This index was the TSE 300 index before 2002. We use this second capital ratio to calculate the capital buffer as the difference – SIZEj. Therefore. book value. compute and to interpret. 2001. the end of that year and was enforced in 1998. N(. We expect a positive relation- the long-term Canadian government bond yield and the T-bill yield. as in Flannery and Rangan (2008). Ux. (2006) and Pathan (2009) as follows: Rj.

526 14.026 0.033 0. VARIABLES DEFINITIONS OBS MEAN STD. 13 In general.258 CAPL Inverse balance sheet leverage ratio = shareholders’ equity/total assets 675 0.009 0.12 We performed a synchronisation be- Our focus on changes in bank capital buffers. the Canadian Imperial Bank of Com- 2. The quarterly average information technique but does not account for ROA is 0. Our sample is composed of the six largest Canadian char. The best case is Allen and Liu (2007).829 BUFR Regulatory capital buffer = CAP  minimum regulatory capital 632 0. For the capital-to-RWA ratio before 1988.2%) and QTOB (16.007 0. Econometric issues merce (CM) and the National Bank of Canada (NA). the Bank of Nova Scotia (BNS). we used the ratio of capital to assets.006 0. 26 subsidiaries of foreign banks and 22 11 branches of foreign banks offering a full range of financial services. Since GMM is an instrumental variables method. and (2) our database has a small TRISK (5.153 0.039 CV Logarithm of charter value 644 21.204 CR3 Concentration ratio = total net income of 3 biggest banks/total net income of all banks 684 0. D’Souza and Lai last quarter of 2010.008 0.040 0.255 3.47%.521 13. ranked by asset size as of the (24 observations). but the lag is usually short. The aver- than two-stage least squares method. risk Compared to other studies of Canadian banking. we use the first Bank of Canada. BUFL is negatively correlated with differences and levels of variables.49%. As of 31 December 2010. equity risk (TRISK) is 1.93%. / Journal of Banking & Finance 37 (2013) 3373–3387 Table 1 Descriptive statistics of the variables (quarterly data from 1982 (Q1) to 2010 (Q2)).19%). we observe an average leverage capital buffer since first differences of variables are all stationary in this context. Finally.09% and an Furthermore. we take one quarter as the lag.018 0.103 36.573 0.008 0. generalized method of moments (2SGMM) estimation technique in For all bank-specific variables. thus reducing the likelihood of spurious regressions From Table 1.055 0.99%.92%.002 0.691 TERM Interest rate term premium = long term government bond yield minus T-bill yield 655 0.398 1. we reports. Data and descriptive statistics that the relationship between bank capital buffers and their perfor- mance may depend on the performance metric used. only used annual data between 1965 and 1989 tered banks.036 0.068 0.45% and idiosyncratic risk (IRISK) is 1.492 0.052 GDPG Quarterly growth rate of real gross domestic product (GDP) 714 0. In fact. Since we are using quarterly data.013 RET One quarter mean equity return based on daily observations 686 0.109 0.156 0.331 0.051 0. BUFL is negatively correlated to the three measures of banks main reasons for this choice: (1) the system of endogenous equa- risks but the correlation coefficients are very low.097 1. competition among Canadian banks. risk and tween market and accounting data as in Claessens et al.46%).044 0.142 BUFL Leverage capital buffer = CAPL  (1/leverage cap) 692 0. These banks ac- count for approximately 90% of the total assets of Canada’s banking In estimating our simultaneous equations.211 ROA Return on assets = net income/total assets 670 0.073 CAPE Economic capital ratio = VaR economic capital/risk-weighted assets 625 0.008 0. 1) of errors in a multifactor model 654 0. who tested of assets. instruments to account for the simultaneity of capital buffers.047 0.009 0. 10 There is therefore no need to use a Haussmann test in our analysis. we have used data from Bloom- a panel data context to deal with potential endogeneity between berg.584 0.92%).34%). Guidara et al.015 0. in absolute value tions that we have chosen is mixed. .043 0. a regulatory capital buffer BUFR of 5.035 0.402 3.006 0. only six.923 2.041 0.007 0. Nathan and Neave (1989) used 39 observations. which is also a limited age quarterly stock return (RET) is 3.269 VTSX Volatility of S&P/TSX index based on daily observations of one quarter 684 0.014 0. means and standard devia- Bond (1998). observations.11 For Canadian economic variables. and to tackle potential serial correlations.13 variables.000 0. Canada’s banking sector is comprised of 22 Canadian banks. DEV Min Max CAP Book capital ratio = GAAP book capital/risk-weighted assets 609 0. We also include lags of each dependent variable as tions) of our sample of quarterly data from 1982 to 2010. supplemented by data collected manually from the banks’ variables. BUFL of 0. Table 1 defines the variables and presents descrip- differences of the dependent variables as done by Blundell and tive statistics (number of observations.002 0. who used 480 quarterly observations. Shaffer (1993).040 IRISK Idiosyncratic risk = conditional variance estimated by a GARCH-M (1.015 0. The banks in our sample. accounting data delay behind market data. our number of and performance adjustments. are as follows: the Royal Bank of Canada (RY). suggesting 3.034 BUFE Economic capital buffer = CAP  CAPE 588 0.3376 A.005 0.000 the Bank of Montreal (BMO).449 0.001 28.018 0.57%) and with ARISK (6.10 BUFL is negatively correlated to RET (12.272 QTOB Tobin’s Q = equity market value/equity book value 645 1.45. observations is substantial.012 0.941 CREDIT Credit ratio = total loan/total asset 675 0.000 1.047 0.50% for the six banks.087 ARISK Implicit asset volatility computed using Ronn and Verma (1986) approach 629 0. with IRISK (3. (2004) used 125 quarterly observations and Gueyie and Lai (2003) used 115 annual the Toronto-Dominion Bank (TD). the fixed effect panel estimation has been favored to the Implicit asset volatility risk (ARISK) is 0. we use the two-step sector and 75% of the assets of deposit institutions. There are two ables. The entire sector managed approximately CAN $3100 billion worth 12 We used over 650 quarterly book observations.023 OUTGAP Cyclical component of the logarithm of real GDP using HP filter 720 0.000 0.077 0.4.000 DREG2 Dummy variable equals 1 over 1998–2010 and 0 otherwise 744 0. as in Flannery and Rangan (2008).006 0. but is positively correlated to ROA (3. (1998) performance lead to the use of only first differences as dependent and Easton and Gregory (2003).000 1. the two-step GMM with panel data is more efficient economic capital buffer BUFE of 3.20% and average Tobin’s Q (QTOB) is 1.021 0.536 0. first difference estimation option which is the only available option Table 2 presents the matrix of correlations between the vari- for the two-step GMM in the panel data context.035 0.600 7. number of banks. we obtained data from use the level and the first differences of the variables as instru- various publications and other sources at Statistics Canada and the ments.139 0.016 0.357 0.471 0.179 TRISK Total equity risk = standard deviation of daily returns over the last quarter 672 0. Quarterly total heteroscedasticity.877 DREG1 Dummy variable equals 1 over 1988–1997 and 0 otherwise 744 0. in that it considers both first not more than 6.411 0.

A. over the sample period. when Basel I rules recent subprime crisis. These findings suggest that Canadian banks Canadian banks’ capital buffers run counter to business cycles? are well-capitalized.007 0. 5 and 6) and the dummy shaped the 2000s. and (iii) an eco- sures are positive: 64.135 1 TERM 0. ing to all of the measures used.109 0.488 0.069 0.162 0.044 0. (2006).009 1 QTOB 0.006 0.160 0.036 0. for the six Canadian banks.041 0. and 10.045 0.015 0.001 0.000 0. These three measures of banks’ capital We use information about business cycles to create three data buffer represent different capital requirement dynamics and one panels: (i) an unconditional panel that considers full business cy- should be cautious in interpreting and generalizing results ob. / Journal of Banking & Finance 37 (2013) 3373–3387 3377 Table 2 Pairwise correlations between banks’ specific variables. Office of the Superintendent of Financial Institutions (OSFI) In the case of BUFL.024 0. when calculating aggregate BUFE. given in tween the variables forming our system of equations.43%). see Saunders right tail of the distribution and use the 95% of the data that remains to calculate the et al.284 0.153 0. Do Canadian banks’ capital buffers run counter to business cycles? atively correlated with BUFE (6. The correlations between the risk mea.22 0.018 0. as described previously. cles without distinguishing troughs from peaks.029 0.236 0. changes in amendment to the Basel I Accord. mean for the banks.038 0.016 0. We use these capital ratios to calculate and is negatively correlated with ARISK (6.058 0.050 0. (2) 1988–1997. after OSFI adopted the 1996 changes in regulatory frameworks. For each panel.165 0. BUFL is positively related to BUFR (1.196 0.024 1 RET 0.101 1 VTSX 0. between ARISK and TRISK.026 0.037 0.341 0. Nor- To answer our research questions.041 1 BUFL 0. expansion panel.141 0. that only considers periods in the trough.1.108 0.008 0.148 0.140 0.134 0.325 0.242 0.377 0.015 0.023 0.142 0. we way and Germany.12% between IRISK and TRISK. since economic capital can be viewed as the level of capital that banks have to hold to remain technically viable (Kretzschmar et al.81% nomic recession panel.124 0.092 0.032 0. Canadian regulations were stricter than regulations elsewhere.270 1 OUTGAP 0.010 0.31%.034 0.033 0.15 show that on average. 2 and 3 of the table present results without controlling for 10% in 2000.068 0.391 0. Canadian banks appear to build up ital regulation? (3) How sensitive is the risk held by Canadian their capital buffer during boom periods and consume them during banks to changes in their capital buffer? (4) How do induced troughs when they are most needed.024 0.007 0. when Canada’s ment between capital buffers (BUFL and BUFR) and business cycles. the associated capital buffers BUFR.086 0. Descriptive statistics for each economic phase.040 0.102 0.015 0.916 0. Guidara et al.076 0.222 1 Meanwhile. Empirical results ernment safety nets.36%) and IRISK (3. which introduced market risk regulatory regimes are controlled with the dummy variable DREG1 as a distinct risk category.060 0.151 0.013 0.087 0.133 1 IRISK 0.067 0. (ii) an economic tained with each measure.001 0. BUFL and BUFE. which sug- multicolinearity in this study is thus very low. The results are presented in Table 4.063 0.174 1 CV 0.043 0.136 0.025 1 TRISK 0. as seen in Table 2.033 0.138 0.093 0.054 0.067 0.037 0.860 0. They also hold less capital in recessions accord- (2) Are Canadian banks’ capital buffers sensitive to changes in cap. gests that Canadian banks hold more capital than what is ‘‘econom- ically’’ required.063 0. 8 and 9).248 0.015 0.667 0.020 1 BUFR 0.172 0. As mentioned above.488 0. The graphs suggest a positive co-move- guish three regulatory regimes: (1) before 1988.003 0.077 0. Lind- banks? quist (2004) and Stolz and Wedow (2011) for the United States.119 0. the relationship appears to change during the adopted the Basel I Accords.211 0. 10. and (3) 1998–2010.49% and the economic Let us look back at our four research questions: (1) Do capital buffer is 3.346 1 CAPE 0.055 0.044 0. Canadian regulations mandated an simultaneous Eqs. respectively. To prevent 15 more bank failures.055 0.687 0.131 0.090 0.040 0.14 These regulatory changes are captured by the variable DREG2 for the 1996 amendment of the Basel I Accord as dummy variables DREG1 and DREG2. Col- 8% minimum capital-to-RWA ratio in 1988.003 0.031 0.031 0.641 1 ARISK 0.134 0. This finding differs from what changes in capital buffers affect the performance of Canadian have been found in other studies such as those of Shim (2013).140 0.081 0.063 0.060 0.610 0.062 0.179 0. The graphs in Fig. the average regulatory capital buffer BUFR is 5.146 1 CR3 0.143 0.081 0. were enforced and the risk-weighted assets (RWA) approach based We further our analysis with a multivariate analysis using the on credit risk was introduced.167 0.069 0.014 1 BUFE 0.061 0. and when the spirit of the Basel II Accord for OSFI’s adoption of Basel I (columns 4.09%.50%.10% between ARISK and IRISK.063 1 ROA 0. this figure rose to umns 1.086 0.033 0.025 0.157 0.020 1 GDPG 0.169 1 CREDIT 0.028 0.070 0.197 0.046 0.170 0. (1)-(3).451 0.049 0.183 0.166 0.025 0.013 0. well as adoption of Basel II (columns 7.108 0. 4.289 0.025 0.30%).010 0. Columns 10. CAPL and CAPE RET is positively correlated with TRISK (5. we exclude 5% of the For a history of Canada’s banking system and changes in regulation.149 0. The risk of Table 3.038 0.056 0.098 0.056 0. Over the sample period of 1982–2010.497 0.043 0. 2010) in a fully disciplined market without gov- 4.045 0. 1 plot capital buffers and the business cycles dian real GDP.023 0.192 0. .162 0. CAP CAPL CAPE BUFR BUFL BUFE ROA RET QTOB TRISK IRISK ARISK VTSX CR3 TERM GDPG OUTGAP CV CREDIT CAP 1 CAPL 0.017 0. that only considers peak periods.071 0. there are no strong correlations be.067 0.135 0.009 0. The correlation between BUFR and BUFE is 15.005 0. but neg.283 0. we distin. we calculate the capital ratios CAP. Finally. Also.31%).070 0.122 0. the leverage capital buffer BUFL is 0.024 0.020 0.080 0.050 0. To alleviate outliers’ bias. CAP is higher than CAPE.568 0. constructed business cycles using the cyclical component of Cana.064 0.07%).456 0. 11 and 12 include both dummies DREG1 and DREG2 and control for 14 Many regulatory in Canada resulted from bank problems in the 1990s.

with the 1996 amendment. Our finding of positive co- explained before. we analyse the impact of regulatory vari- force banks to lend less during recessions. this situation may to regulatory changes.000 study period. and some claim that it helped make Canada’s the two regulatory regimes. after 2000. cyclical by design. Also. when it rose to 23 un- ceiling on an unweighted leverage ratio.007 Non-conditional 0. These regulatory changes increased the level of capital in Can- capital during boom periods and consume a portion of this buffer ada’s banking sector after 1998.06 regulation? 0. In- (HP) filter to obtain cyclical components. In fact.06 requirement rose to 10% in the beginning of the 2000s. (A and B) Banks’ capital buffer and business cycles in Canada. meeting the capital requirement.048 0.BUFR and business cycles 0. 0.074 0. As and potentially aggravating a downturn. leading to credit crunches ables DREG1 and DREG2 on the change in bank capital buffers. To ital ratio.005 Economic capital is calculated using Value at Risk (VaR) at the 99. the ratio fell.01 Basel II. This period corresponds with enforcement of the advanced 0. Guidara et al. Business Capital ratio Economic capital Inverse leverage Regulatory capital Economic capital Leverage-based cycles (CAP) ratio (CAPE) ratio (CAPL) buffer (BUFR) buffer (BUFE) capital buffer (BUFL) Expansion 0. The leverage-based capital buffer (BUFL) is equal to the difference between the shareholders’ equity-to-assets ratio and the inverse of regulatory ceiling on an unweighted leverage ratio. compute this last variable.03 -0. After Fig.02 -0. from 1893 to 1982.097 0.035 0.g. creased the minimum regulatory capital-to-RWA ratio from 8% to We interpret this result as evidence that Canadian banks build up 10%. Are Canadian banks’ capital buffers sensitive to changes in capital 0.04 0.047 0. Unfortunately.06 of the amendment of Basel I in 1996.02 0.002 Fig.008 -0. 3. we observe sudden increases. 2009.02 0.139 0. especially after the announce- ment of the Basel I Accords in 1987 and after the announcement 0. The regulatory capital buffer (BUFR) is defined as the difference between banks’ capital ratio and minimum regulatory capital ratio.BUFL and business cycles 4.00 -0. where it remained until 2000.3378 A. we first adjust the seasonal components of capital buffer Canadian regulators not only maintained the minimum regulatory (by using the moving average over four quarters).16 The increase is more pronounced in the periods leading up to regulatory change. The graphs confirm our previous -0.002 Fig. 3 plots the banks’ average capital-to-RWA ratio over time. As for the impact of Basel II 0. since one would have expected the in the lean times. -0.. changes in regulatory regimes are controlled movement between capital buffers and the business cycle. 2 shows business cycles and regulatory regimes over the 0.102 0. even after the minimum regulatory capital -0.97% confidence level.06 Fig. 2009). Blum. 5 and 6 of Table 4 for the results) and DREG2 for the dian banking sector. but also reduced the leverage ratio limit.051 0. Canada in- fers (DBUFLt) and the output gap (OUTGAPt) over the sample period. The regulatory capital buffer (BUFR) is defined as the difference between banks’ capital ratio and minimum regulatory deed.04 sets divided by shareholders’ book equity. The left hand axis represents values of the cyclical part of capital buffers.02 took place in 2008. The right hand Canadian regulators adopted the Basel I regulations and banks be- side axis gives values of business cycles measured by the cyclical component of real gan to account for their credit risk in the denominator of their cap- GDP.04 B. The leverage-based capital buffer (BUFL) is equal to the difference ratio at 30 from 1982 to 1991.04 since the early 1980s.00 credit risk approach in Canada. BUFL CYCLE Thus.g. 2008).004 along with the balance sheet leverage ratio.005 Recession 0.02 -0.04 on Canadian banks. In the next section. 1. then we use the Hodrick–Prescott capital requirement. A.02 claim about the adequate level of capitalization of Canadian banks -0.006 -0. their average capital-to-RWA ratios increased over the study 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 period after a secular decrease in banks’ capital. Canada’s banking authority capped the balance sheet leverage capital ratio..032 0. However. This leverage ratio requirement has been shown to mitigate asymmetric information and agency problems (e.03 0.047 0. der certain conditions. 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 no risk-adjusted capital ratio requirements existed prior to 1988 be- BUFR CYCLE cause Basel I guidelines were only introduced in that year.004 0.. / Journal of Banking & Finance 37 (2013) 3373–3387 Table 3 Aggregate capital buffer measures.006 0. 4 years after OSFI started progressive adoption of 0. Dickson. measured by total as- -0. The economic capital buffer (BUFE) is defined as the difference between banks’ capital ratio and their economic capital ratio. however. First.051 0. we observe that the increase in the capital ratio 0. The explanations for these observed trends are as follows. . we examine whether the countercyclical ef- fect found above is sensitive to changes in the regulatory 16 Saunders and Wilson (1999) document a continual decrease of banks’ capital environment.00 0. Late in 1991. in that it requires banks to increase their capital To determine how sensitive Canadian banks’ capital buffers are ratio when they face greater risks. the authority decreased between the shareholders’ equity-to-assets ratio and the inverse of the regulatory the limit to 20. As shown in panel A of -0. The results for each model show a po- banking sector more resilient to the recent credit turmoil (e.044 0.138 0.01 -0. with dummy variables DREG1 for OSFI’s adoption of Basel I (see can be interpreted as evidence of countercyclical effects in the Cana- columns 4. many capital ratio to decrease or remain more or less the same after the critics have pointed out that the Basel capital regulations are pro- introduction of market risk as a new risk category.2. sitive and significant relationship between variations in capital buf- Bordeleau et al.098 0.139 0.

005) (0.056) (0.850)  DBUFL DREG2 0.211 0. The values in parentheses are robust standard deviations.050) (0.064 0.007) CV 2.0038 0.0091) (0.139 0.673 17. The estimations are performed using two-stage GMM regressions for panel data (2SGMM).0006 0.160) (0.0114 0.575) DREG2  OUTGAP 0.0009 0.186 0.0013 0.0021*** 0.016) (0.0025*** 0.0154** 0.25 (0.180*** 0.591*** 3.0504 0. we use total equity risk (TRISK) as the risk measure and equity returns (RET) to measure performance.0003) (0.007) TERM 0.040) (1.0029 0.0002) (0.41 (0.0299 0.166*** 0.055) (0.0009) (0.006) (0.308 (0.683) (13.087 0.0009) (0. / Journal of Banking & Finance 37 (2013) 3373–3387 DREG1 OUTGAP 0.0017* (0.0091) (0.389) (0.0041 0.099) (0.0415*** 0.0006*** 0.0003) (0.385) (0.0004) (0.0014*** 0.0010) VTSX 0.242 0.0005*** 0.00004) CR3 0.002) (0.835 0.105 0.003) (0.113 0.054) (0.646) (0.137 0.007) (0.012) (0.0012 0.007) (0.267 3.112 0.468 0.088 0.248 0.0183)  ** A.46E05 6.0258* (0.010) (0.0212 (0. ** 5% Significance level.0093) TRISKt1 0. model 3 in columns 7–9 and model 4 in columns 10–12.0013*** 0.399 14.237) (0.0144) DREG2 0.069) (0.015) (0.054) (0.780) SIZE 0.0053 0.468) (0.210) (0.0012*** 0.0012*** 0.050) (1.007) (0.0006) (0.0390*** 0.0003 0.0118 0.188*** (0.009) (0.105 0.005) DREG1 0.0003 0.0066 0.052) Observations 565 565 565 565 565 565 565 565 565 565 565 565 R-squared 0.673 (0. 3379 .003) (0.524) DBUFL 0.334 0.0009 0.573) DBUFL  DREG1 0.002) (0.165 0.0015 0.008) (0. model 2 in columns 4–6.00005) (0.49E05 (0.0034 (0.0366*** 0.859 0.0002) (0.004) (0.048) (1.045 0.214) (5.0099) (0. Financial data are quarterly observations from statements covering 1982–2010.528 (0. Guidara et al.0128* (0. *** 1% Significance level.0103 0. The capital buffer (BUFL) is calculated as the difference between the shareholders’ equity-to-assets ratio and the inverse of the regulatory ceiling on an unweighted leverage ratio. Model 1 is shown in columns 1–3.0009) (0.270) (0. * 10% Significance level.0345*** 0.0003) (0.0025 0.078 1.0844 1.175 0.006) (0.0001) (0.011) (0.0569*** (0.0115) (0.200) (5.031) (0.0031* (0. risk (TRISK) and performance (RET).135* 0.0817** 0.007) (0.045) RETt1 0.0094 0.0022** 0.240) (3.0078 0.004) (0.0009 0.017) (0.092 Number of banks 6 6 6 6 6 6 6 6 6 6 6 6 This table presents regression results of systems of simultaneous equations of changes in bank capital buffers.087 0.0040** 0.0296*** 0.326*** (0.005) (0.0006) BUFLt1 0.314) (0.459) (0.0098 0.180*** 0.011) (0.031) (0.143 0.0599 0.0056) (0.630) DTRISK 0.0597*** (0.005) (0.226) (3.086 0.677) (0.0567*** 0.321) (0.56E05 6.012) (0.0395 0.005) (0.81 (0.0008*** 0.709) (13.0007) (0.0037) (0.0008*** 0.77E05 1.0032) (0.0202 0.0087) (0.201 0.0027 (0.23E05 7.014) (0.602 0.0007 0.Table 4 Estimation results of the simultaneous equations of variations in regulatory capital buffer (BUFL).0084) BUFR 0.53 15.002) (0.0006) (0.0060* 0.0311 0.0001*** (0.411) DRET 0. Market data was extracted from daily data and converted into quarterly units.45 (0.0270*** 0.0169 0.0002) (0.0007) (0.168) (0.0068* 0.0100* 0.00004) (0.704) (13.0594 0. risk and performance.00004) (0.051) (1.0154 0. All other variables are defined in Table 1.149*** 0.0184*** 0.0075 0.0002) (0.0002 2.0018*** 0.510** 0.0397*** 0.441) (0.140) (2.033) (0.0043 0.002) CREDIT 0.0503*** 0.0007) (0.82E05* 0. In this table.0003) (0.016) (0. Model 1 Model 2 Model 3 Model 4 (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Variables DBUFL DTRISK DRET DBUFL DTRISK DRET DBUFL DTRISK DRET DBUFL DTRISK DRET OUTGAP 0.

06 0 -0. How sensitive are Canadian banks’ risk to changes in their capital ratio and has been taken from Canadian banks’ official publications. The In the second graph (panel B). Canadian banks’ risk.00 -0.02 0.02 -0.25 1 the Basel II regulations. Hence. in 1988.04 0. the Canadian banking ital buffer. while the dummy DREG2 has a positive significant impact on BUFL.Capital/RWA and asset/equity during the 1996 amendment and Basel II periods. market idiosyncratic risk) and ARISK (the implicit volatility of the banks’ assets).05 ulatory changes probably helped boost the capital base of Canadian 10 0. the capi- tal-to-RWA ratio was increased from 8% to 10%. For the maximum leverage ratio (MAX_LEV).Capital/RWA Basel I regulations and then the 1996 amendment of Basel I and 0.04 -0.04 0.02 -0. the bal- 25 0. and after 1988 it is computed as the capital-to-RWA 4. Canadian banks have . 4 depicts the pattern of Canadian banks’ equity risk (TRISK). products OUTGAP  DREG1 and OUTGAP  DREG2.08 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 REG CYCLE TRISK CYCLE VTSK Fig.10 the business cycle is not affected by the two Basel regulatory re- 0. Indeed.06 -0. This relationship between business cycles and capital buf- 0. In buffer? 2000. the magni- tude of the positive co-movement between capital buffers and 0.00 0. over the entire sample per- 1996 amendment of the Basel I Accords and the adoption of Basel II iod. since many authors have criticized the Basel Accords as being procyclical 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 by design. the scale on the right hand axis is for the capital ratio relationship between TRISK and business cycles is ambiguous. The two graphs show the average ratios for the six big chartered banks. once again we turn to our system of simultaneous equa- supervisory authority fixed a balance sheet leverage ratio cap of 30 from 1982 to tions. However.06 1 0. (CAP) measure and the left hand scale is for the balance sheet leverage ratio To address the sensitivity of banks’ risk to changes in their cap- measure (LEV). and (3) the period in the spirit of Basel II starting in 2004. where columns 1–3 are for TRISK.10 1991 and was increased to 23 in 2000. the effect is less pronounced during the business cycles and regulatory environments by using the cross.00 0 fers during the Basel Accord periods is very interesting.04 0. 2011). the limit was decreased to 20 and this was the ceiling until 2000. 4. 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 Having studied the behavior of Canadian banks’ capital buffers during different business cycles and changes in capital regulations. Up to 1988. the regulatory dummy DREG1 alone has a significant 30 0. but are positively and significantly related to varia- regimes in the regressions. We control for the combined effects of tions in ARISK.25 However. 0. To compute the last two variables.3380 A. The gray areas designate major Fig. We use two additional risk measures: IRISK (the banks’ 1991.g. (2) implementation of the 1996 amendment of Basel I taking effect in 1998 The right hand axis represents the levels of average banks’ equity risk (TRISK) and and (3) the period in the spirit of Basel II starting in 2004. the minimum capital ratio was increased to 10% in Canada. These capital reg- 15 0.08 0. the enforcement of Basel I by the Office of the Superintendent of Financial Institutions (OSFI).00 0.06 0. We find that. especially after the 1996 amendment and during the REG CAP MIN_CAP Basel II period (e. after 2000. 1988. Also.04 -0. columns 4–6 for IRISK and columns 7–9 are for the ARISK risk measure.20 The relationship between variations in BUFL and DREG1  OUT- 0. respectively.02 0. 1996 amendment and Basel II periods.04 -0. Business cycles and capital regulations. In columns 10. Jokipii and Milne.15 GAP and DREG2  OUTGAP are not significant. the bank’s capital ratio was computed as the bank’s capital-to-total assets ratio. Thus. market risk and business cycles. 3. we include the dummies for both regulatory risk (IRISK). we first component of real GDP. for A. / Journal of Banking & Finance 37 (2013) 3373–3387 0. 2008. then we use the Hodrick–Prescott (HP) filter to obtain cyclical parts. changes in banks’ capital buffer BUFL are not significantly re- (see columns 7. The results are presented in Table 5.3. We instead find a countercyclical effect during Basel I as well as B. adjust their seasonal components using a moving four-quarter average. 8 and 9 of Table 4 for the results). we consider a minimum capital ratio (MIN_CAP) of 8% as fixed by Basel I in 1987. Late in 1991. serve weak co-movement between VTSX and business cycles. (2) implementation of the 1996 amendment of Basel I that takes effect in 1998.12 0. Before sures of Canadian banks’ risk. lated to variations in equity returns risk (TRISK) or idiosyncratic 11 and 12 of Table 4. Guidara et al. 35 0. We ob- enforcement of Basel I in 1988.00 banks. 2013. The gray areas designate Fig. 2.15 ance sheet leverage ratio limit was decreased from 30 to 20 in 20 0. The left hand axis regulatory changes in the Canadian banking sector: (1) enforcement of Basel I in gives values of the business cycles measured by the cyclical component of real GDP. CYCLE is the cyclical Canadian market equity risk (VTSX). when it was increased to 23 under certain conditions. Trend of banks’ capital and leverage ratio between 1982 and 2010. major changes in capital regulation in the Canadian banking sector: (1) the Canadian stock market risk (VTSX) and business cycles. MAX-LEV LEV CAP we now turn to the impact of changes in capital buffers on mea- Fig.20 negative impact on the variations of BUFL.05 gimes.. Shim.

Model 1 is shown in columns 1–3.188*** 0.005) (0. This may be seen as evidence of the success of the the positive co-movement between capital buffers and the busi- combination of risk-based capital requirements suggested by Basel ness cycle is present for all risk measures used.0235 (0.011) (0.000) DBUFL  DREG2 0.0237 0.0066 0.589) (0.0012 (0.0855 0.81 0.0018) (0.909*** (0.0031* 0.102) (0.0144) (0. ** 5% Significance level.685) DBUFL 0.0183) (0.46 0. Also.0059** 0.399 14.0007 0.012) (0.0013*** 0.0153) DREG2 0. The estimations are performed using two-stage GMM (2SGMM).0010) VTSX 0.014) (0.0033* 0.0014 0.718) (13.0005*** (0.0035* 0.095) (0.0114 0.689*** (0.0018) (0.0446 0.051) (1.011) (0.25 (0.26E05 (0.007) (0.211 0.544 0.021) (0.0009) (0.017) TERM 0. we find that any of the risk measures.77 4.115) (0. *** 1% Significance level. better succeeded at aligning risk-taking and their capital base.0258* 0.012) (0.0077 0.19 (0.810) DTRISK 0. Canadian regulators.007) (0.033) (0.0038 0.031) (0.91 4. sponse to risk exposures.709) (13.0128* 0.0345*** 0.092 0.53 15. this Accords and the non-risk-based capital constraints imposed by is consistent with the second pillar of Basel II on maintaining a per.0556*** 0.539) (0.018) (0.0009 0. risk and performance.704) (13.524) (0.548 (0.0025) (0.0017* 0.019) (0.0002* 0.0597*** 0.411 (0.002) (0.189*** 0.0241* 0.0003) (0.017) (0.0344*** 0. / Journal of Banking & Finance 37 (2013) 3373–3387 3381 Table 5 Estimation results with different risk measures.586) (0.00004) (0.0275* (0.0093) (0.0283 0.0007) (0.0044* 0.326*** (0.673 17.0748 0.263 14.0155 (0.0009 0.0074) (0.729) DBUFL  DREG1 0.052) (0.0097 0.308 0. Market data was extracted from daily data and converted into quarterly units.0027) (0.016) (0. we use three risk measures: total equity risk (TRISK).0003 0.0009 0.080) ARISKt1 0. idiosyncratic risk (IRISK) and implicit volatility of assets (ARISK).0299 0.109 0.712) DREG2  OUTGAP 0.0144) (0.108*** 0.0866 0.045) IRISKt1 0.0035 0.134) DRET 0.085 Number of banks 6 6 6 6 6 6 6 6 6 This table presents regression results of systems of simultaneous equations of changes in bank capital buffers.0004) (0. note that it (see Table 4).741) (14. Values in parentheses are robust standard deviations.0008*** 0.117 16.0020) (0.0003) (0.0084) (0.0095) TRISKt1 0.573) (0.242 0.0328 (0. * 10% Significance level.0034 0.87 (0.850) (1.024) DREG1 0.216 0.0003) (0.66E05 7.0006) BUFLt1 0.0202 0.141 0.0569*** 0.9 (0.0027 0.49E05 2.683) (13.0187 0.54 (0.0006) (0.0020) CREDIT 0.0001) (0.780) (1.054) CV 0. .053) Observations 565 565 565 565 565 565 556 556 556 R-squared 0.0170) (0.0019** 0.0006) (0.411) DIRISK 0.41 0.0006** 0.0212 0. amendment to Basel I and Basel II by OSFI.0019) (0.22 (0. A.868) (15.190*** (0.47E05 0. Financial data are quarterly observations from statement covering 1982–2010. we use equity return (RET) to measure performance. Model 1 Model 2 Model 3 (1) (2) (3) (4) (5) (6) (7) (8) (9) Variables DBUFL DTRISK DRET DBUFL DIRISK DRET DBUFL DARISK DRET OUTGAP 0.667 0.0001) CR3 7.45 0.0076) BUFR 0.778) (14. Guidara et al.0008 0.0012*** 0.0002) (0.0012*** 0.574 17. The capital buffer (BUFL) is calculated as the difference between the shareholders’ equity-to-assets ratio and the inverse of the regulatory ceiling on an unweighted leverage ratio.004) (0.016) (0.673 0.710) (2.0003) (0.99 4.0111 (0.0004) (0.143 0.830) (2.0010) (0. The relationship between risk and OUTGAP is not significant for Analyzing the regulation dummies separately.0009*** 9.724* 15.528 0.014) (0.0003) (0.386) DARISK 0.0022 (0. The co-movement between business DREG1 positively impacts TRISK and that DREG2 negatively affects cycles and measures of risk thus does not appear.950) (2. model 2 in columns 4–6 and model 3 in columns 7–9.700* 18.575) (0.0102** 0.0011*** 0.673) (13.0091) (0.184 13.930) SIZE 0.002 0.117) (0.0007) (0.095) (0.0078*** 0.045) RETt1 0.0032 (0.0004) (0. In this table.051) (0.0013*** 0.0533 (0.0500*** (0.0177) DREG1  OUTGAP 0.0001*** 0.083 0.0035 0.733) (13.105) (0.031) (0.003) (0. All other variables are defined in Table 1.0309 0.630) (1.355 0.0014*** 0.011) (0.003) (0. especially after the adoption of the 1996 manent supervisory review process to adjust capital levels in re.

has a positive sig- effect of the buffer on ROA remains positive. (2012).3382 A. we can connect the impact of these two variables to the findings of Antia et al. The results indicate that. we once again turn to our system of simultaneous The subprime crisis may introduce biases in our results because equations. It can be interpreted as Canadian -0. In fact.17 The positive co-movement between capital buffers and the Since variations in both capital buffers and in banks’ risk-ad. / Journal of Banking & Finance 37 (2013) 3373–3387 0. (1999) and Balli et al. Columns 4–6 of Table 7 give the results for BUFR and columns 7–9 for BUFE. The results available in columns 4–6 of Table 8 confirm that tory periods. having smoothing the subprime crisis period which begins in 2007. this implies that Basel I and Basel II have im- 0. -0. we observe that positive variations in ital buffer and absorbed it in difficult times. columns 4–6 for return on assets order to study the sensitivity of the results to the subprime crisis (ROA) and columns 7–9 for Tobin’s Q (QTOB) as performance we perform our regressions excluding the subprime crisis period measures. Second. it is significantly negative. between business cycles and capital buffers. measured by the cyclical component of real GDP. sults. in the next (see Table 5) and performance (Table 6) used. confirming the results of selecting a relevant measure of capital buffers.00 founding effects. the impact is not significant. as was the case for the entire sample. the volatility of the Canadian on assets. Fig. in part. VTSX and TERM can be seen as capturing. The conflicting results obtained from ritization in the late 1990s enabled banks to hedge the market risk this metric can be explained by difficulties obtaining comprehen- component of their portfolio. and credit mar. capturing shocks on the term structure Q means that banks tend to reduce their level of investment when of interest rates. we first adjust the seasonal components of RET using the impact on changes in return on equity of Canadian banks. After controlling for other con- 0. securitization. The results are presented in Table 7. How do induced changes in capital buffers affect the performance We now turn to the sensitivity of the results to other capital of Canadian banks? buffer measures. 5. then we use the Hodrick Prescott (HP) filter to obtain they negatively impact ROA and positively impact Tobin’s Q. At the same time. the introduction in 1998 of market risk as a distinct risk category pushed banks to reshuffle their asset portfolios to- wards assets with low market risk charges. This result highlights the importance negatively correlated to the business cycle.06 The regulatory environment dummies DREG1 and DREG2 have a 0. 5. But changes in banks’ capital buffers are still not significantly related to the change in their equity 17 We thank an anonymous referee for bringing this to our attention. the reg- capital buffers are not a significant factor explaining variations in ulatory regimes have not had specific effects on this co-movement banks’ return on equity. (2007–2010) to check if the crisis had a special effect on our re- For RET. the development of market derivatives and credit secu- risk-weighted assets (RWA). for exam- of Table 2.1. TERM.06 We also analyze the interactions between capital buffer and regulatory dummy variables in the same equations (DREG1  1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 DBUFL and DREG2  DBUFL) to account for the effect of capital RET CYCLE buffer on performance following the adoption of Basel I and Basel II. by capital markets. respectively. business cycle observed in Table 4 holds for all measures of risk justed return on capital and hence their performance.04 positive and significant impact on banks’ equity returns (RET). In 1–3 are for equity returns (RET). a proxy for Canadian market risk. while moving four-quarter average. for all of the performance measures. Guidara et al. More precisely. returns risk. Furthermore. As for the . Finally. as 0.2. section we analyze how changes in the capital buffer affect banks’ performance. effects on banks’ risks. It appears that bank equity returns are icant for either BUFR or BUFE. Fig. 5 depicts Canadian banks’ performance measured by equity The effect of the business cycle on bank capital buffers is not signif- returns and business cycles. In this section we verify the sensitivity of our results to different kets.08 -0.02 banks being well-capitalized and able to easily meet Basel capital -0. federal tax-transfer systems. the combined negative impacts of the buffer and regulation significantly reduces the variation in return Moreover. With the other performance measures. The results are presented in Table 6. To I and Basel II. These cyclical components.00 -0.04 0. it indicates that bank buffers have no compute the last variable. Robustness checks on the Canadian gross provincial products were smoothed. 5. Thus. Excluding the subprime period To determine how capital buffer variations affect these banks’ performance. But the overall market index. to compute BUFR and BUFE. 10% confidence level. with the development of sive data on risk-weighted assets. we need to obtain or compute tors. capital markets model specifications such as different capital buffer measures and and credit markets. First. and for ROA. In fact.02 well as return on assets (ROA). we observe that VTSX. we observe instead that a variation in bank capital buffers has a po. the result for Tobin’s bonds and the T-bill yield.02 0. where columns of the extreme volatility observed in the data during that period.04 requirements. The right the behavior of DBUFL  DREG is more or less the same under Basel hand side axis represents the levels of banks’ average equity return (RET).4. Canadian banks increased their cap- significantly positive. Banks’ performance and business cycles. therefore. this differs from the results across the entire sample. who show that shocks 5. This may be explained by a combination of several fac- ple. the coefficient is not significant over the Basel regula. the results on equity returns suggest that market participants did not difference between the yield on long-term Canadian government find this regulation to affect them. it is in period of economic booms. where significant and negative impact on QTOB. meaning that it reduces bank earnings. The non-significant nificant impact on TRISK and IRISK. The left hand axis gives values of the business cycles.06 -0. while it has a banks. has a negative significant impact on TRISK at the the capital buffer increases during the Basel regulatory periods.06 0. For QTOB. Although not directly related to our work. Alternative capital buffer measures 4. results are intuitive.04 -0. An increase in risk reduces the excess capital buffer held by sitive and significant impact on variations in ROA.02 proved banks’ performance.

0246 41.514 0.157 Number of banks 6 6 6 6 6 6 6 6 6 This table presents regression results of systems of simultaneous equations of changes in bank capital buffers.0031) (0.138*** (0.34) DBUFL  DREG2 0.0071) (0. as was the case for the entire sample.0299 0.0258* 0. The estimations are performed using two-step GMM regressions (2SGMM).0069) (0.0292) (0.0010 0.0744) QTOBt1 0.177*** (0.0058 0.177*** 0. To implement this approach as additional robustness test.0595*** 0.0008*** 0. A.126) DQTOB 0.0124) (0.0009 0.0411) BUFR 0.0011) VTSX 0.464) DBUFL 0.0016*** 0.0018) (0.0003) (0.0828) DREG1  OUTGAP 0.0071) (0.0330) (2.0006 0. the interactions between the regulatory dummies ðDBUFL  DBUFLÞ  DREG instead of DBUFL  DREG.0154** (0.25) DTRISK 0.524) (0.0003 0.0038 (0.0313 0. *** 1% Significance level.725) (66.0003 0.0067) (0.0308) (0.0579) (0.308 0.097 0.575) (0. Market data was extracted from daily data and converted into quarterly units.0066 0.0002) (0.0638) DREG2 0.0530*** (0.0001*** 0.0070) (0. Financial data are quarterly observations from statements covering 1982–2010.0017* 0.0120) (0.0030 0.0085) (0. * 10% Significance level.573) (0.0018) (0.0008) (0.791 0. Since DREG is a .0034 (0.0504) (0.399 14.0042 0.261) (0.326*** 0.211 0. / Journal of Banking & Finance 37 (2013) 3373–3387 3383 Table 6 Estimation results with different performance measures.49E05 4.0037 0. ** 5% Significance level.509 172.0006*** 0.0336) (0.0315) (2.327*** (0.0127) (0. for Models 2.530 15.143 0.0344*** 0.0144) (0.0084) (0.0011 0.3*** (0.0038 0.0020 (0.0001) (0. Guidara et al.528 0.0006 0.0202 0. risk and performance.64E05 (0.673 17.0009 0.63) (0.683) (13.0003) (0.0309) (0.0012*** 0. Values in parentheses are robust standard deviations.194*** (0.0008*** 0.0139) (0.0006) (5.0588*** 0.0514) (1.0019) CREDIT 0. cross products of the variables: OUTGAP  DREG and DBUFL  DREG.0085) (0.0027) BUFLt1 0.666 0.0003) (0.0093) (0.548 (0.0114 0.701 (0.0049) DROA 0.0031* 0.0014 0.0569*** 0.65*** (0.149 0.0003) (0.0048 0.268) (0. are captured by the significant impact.0128* 0.29E05) (4.0004) (0.250 0.702) (64.0076) CV 0.792) DBUFL  DREG1 0.17E05) CR3 7.0261** 0. return on assets (ROA) and Tobin’s Q (QTOB).0168) (0.197 0.748*** 0.0004) (0.0007) (0.0027 0.030 0.13) SIZE 0.0226 0.22) DRET 0.0093) (0.0007) (0.0001*** 0.0459) RETt1 0.711) (0.403 (0.78) (0.81 0.0515) ROAt1 0.0008*** 0.0031* 0.0001*** (4. Model 1 Model 2 Model 3 (1) (2) (3) (4) (5) (6) (7) (8) (9) Variables DBUFL DTRISK DRET DBUFL DTRISK DROA DBUFL DTRISK DQTOB OUTGAP 0.738*** 0.0142) (0.0007) (0.704) (13.0004) (0.0010) (0.0345*** 0.596 171.333 181.709) (13.0066) TERM 0.0009 0.0313) (2.0127* 0.0459) (0.0034 0.0035*** 0.0007 0.0003) (0.2*** (0.0352 0.0306) (0. The capital buffers (BUFL) are calculated as the difference between shareholders’ equity-to-assets ratio and the inverse of the regulatory ceiling on an unweighted leverage ratio.0452) (0. Model 1 is shown in columns 1–3.0327) Observations 565 565 565 565 565 565 565 565 565 R-squared 0.0173 0.0030 0. All other variables are defined in Table 1.166 (0.0137) (0.41 0.728*** (0.0154) (0.715*** 0.547) DREG2  OUTGAP 0.0014*** 0.708) (0.328 0. Recently.65E05) (0.686) (0.0597*** 0.0008) (0.0007) (0.261) (0.0003) (0.0098) TRISKt1 0.0010) (0. we use total equity risk (TRISK) to measure risk along with three measures of performance: equity return (RET).0003) (0.04E05 0.0041** (0. Alternative methodology to capture the interaction terms the variables constituting the interaction terms by their means.0013*** 0.673 0.092 0.27E05) (4. model 2 in columns 4–6 and model 3 in columns 7–9. we con- As in Rajan and Zingales (1998) and others.0017* 2. performance measure RET.0621) (11.85) (0.0156) (0.326*** 0.45 0. variations in capital buffers have no (DREG1 and DREG2) and OUTGAP or DBUFL.292*** (0.0312) (0. Balli and Sorensen (forthcoming) suggest adjusting 5.0230 0.7*** (0.0013*** 0.0009) (0.3.0013*** 7.188*** (0.0183) (0.0134 0.733) (65.0635*** (0.411) (0. In this table.0019) DREG1 0.24E05 0. 3 and 4 sider ðOUTGAP  OUTGAPÞ  DREG instead of OUTGAP  DREG and shown in Table 4.0333) (0.0212 0.

386*** 0.453) DRET 0.0114 0.0343) (0.709) (13.0194 0.365) (7.229*** 15. *** 1% significance level.176** 0.146 Number of banks 6 6 6 6 6 6 6 6 6 This table presents regression results of systems of simultaneous equations of changes in bank capital buffers.780) DBUFR  DREG1 10.0533 0.0062 0. Model 1 Model 2 Model 3 (1) (2) (3) (4) (5) (6) (7) (8) (9) Variables DBUFL DTRISK DRET DBUFR DTRISK DRET DBUFE DTRISK DRET OUTGAP 0.0334 (0.0008** 0. Capital buffers (BUFL) are calculated as the difference between the shareholders’ equity-to-asset ratio and the inverse of the regulatory ceiling on an unweighted leverage ratio.0067 0.884) DBUFL  DREG1 0.001) (0.037 0.49E05 0.0084) (0.956 (0.0013) (0.810) (0.866*** (0.008) CV 0.052) (0.683) (13.0031* (0. .0005 (0. Guidara et al.0004 0.00*** 133.0093) BUFRt1 0. we use total equity risk (TRISK) to measure risk.163) (5.777) (0.0006) BUFLt1 0.007) (0.185 0.0004 1.0109 0.148) (8.067) (0.0083) BUFR 0. and equity return (RET) to measure performance.0183) (0.124 0. / Journal of Banking & Finance 37 (2013) 3373–3387 Table 7 Estimation results with different capital buffer measures.0012*** 0.007) (0.34** 4.228*** (0.81 (0. In this table.016) (0.0794) (0.420) DBUFR  DREG2 11.575) (0. ** 5% significance level.012) (0.495 0.237) (50.0062 (0.004) (0.971 4.0299 0.045) (0. model 2 in columns 4–6 and model 3 in columns 7–9.0017 0.0046) DREG1 0.0239 (0.125) (1.031) (0.0007) (0.0053) (0.0006) (0.199** 0.092 0. Market measures were extracted from daily data and converted into quarterly units.0311 0. * 10% significance level.048) RETt1 0.0159) DREG1  OUTGAP 0.0101 (0.190) SIZE 0.0422*** (0.887 (0.028) (0.0001*** (0.003) (0.986*** 0.0212 0.080) (0.0569*** (0.41 (0.047) (0.308 0. Financial data are quarterly observations from statements covering 1982–2010.0016) (0.2950) (0.189 3.411) (0.263) (3.0009 0.051) (1.00004) CR3 7.0066 0.049) (0.0038 0.0487) (0. Economic capital buffers BUFE are the difference between banks’ actual capital ratio and their economic capital ratio (CAPE).45 (0.188*** 0.014) (0.867 1.524) (0.0042 0.041 1.05*** 134.263) (3.0597*** 0.0713) (0. Values in parentheses are robust standard deviations.0002 0.0008*** 0.035 (0.149) (0.0003) (0.0036 (0.0021** 0.114 0.00020) (0.96*** 133.007) TERM 0.0003) (0.0001*** 0.0176) (0.423 0.0004) (0.0128* 0.185 2.0003 0.0053) (0.0*** (3.192*** 0.03 1.0239* (0.0169** 0.225) (50.0017* 0.0060** 0. risk and performance. The estimations are performed using two-step GMM regressions (2SGMM).143 0.0219 0.621*** (0.013) BUFEt1 0.271) (3.360) DBUFE 0.0009 0.00004) (0.0653*** (0.006) (0.326*** 0.155 (0.0202 0.0128 (0.109 11.871) DREG2  OUTGAP 0.704) (13.0127** 0.086) (0.662 0.033) (0.0677 0.338*** 0.031) (0.349*** (0.0*** (3.196) TRISKt1 0.0521** 0.3384 A.236) (50.821) (0. The regulatory capital buffer (BUFR) is calculated as the difference between the book capital ratio (CAP) and the minimum regulatory capital requirement.0025* 0.0091) DREG2 0.9*** (3.007) (0.0206) (0.0034 0.0024) (0.041) (1.161 3.0172 0.0496) (0.856) DBUFL 0.0632) (0.0004) (0.0171 0.00041 (0.0137*** 0.850) DBUFL  DREG2 0.0699* 0.013) DTRISK 0.630) DBUFR 11.0144) (0.573) (0.0145 0.0345*** 0.53 15.160) (5.0013*** 0.0007) (0.87 (0.530) DBUFE  DREG1 0.0004) (0.002) BUFL 0.022) (0.399 14.528 0.0027 0.129) VTSX 0. Model 1 is shown in columns 1–3.0010) (0.0009** 0.055) Observations 565 565 565 564 564 564 556 556 556 R-squared 0.0258* 0.066) (0.0010) (0.110) CREDIT 0.673 0.0033 0. All other variables are defined in Table 1.211 0.673 17.172) DBUFE  DREG2 0.047) (0.148) (5.005) (0.0007 0.25 1.

033) (0. In this table. after OSFI adopted This paper examines the cyclical behavior of Canadian banks’ the 1996 amendment to the Basel I Accord. the mean is zero by construction dian banking sector in its use of a comprehensive dataset over a ðOUTGAP ¼ 0Þ.0078 0.0006 0.053) (1.007) (0. we are using DBUFL instead of BUFL.0008*** 0.0032 0.014) (0.0319 0.673 0.031) (0.007) (0.031) (0.115 18. we use total equity risk (TRISK) to measure risk and equity returns (RET) to measure performance.0002 0.045) (0.249 0.0007) (0.0003) (0.0013 (0.0013*** 0. when OSFI adopted and en- forced the Basel I Accord.186*** (0.49E05 2. All other variables are defined in Table 1.0114 0. it stands out among studies of the Cana- rick–Prescott filter.002) (0.0597*** 0. and the Basel II period. for the interaction terms involving capital relatively long time frame (1982–2010).399 14.850) (0.510) SIZE 0. / Journal of Banking & Finance 37 (2013) 3373–3387 3385 Table 8 Estimation results excluding the subprime crisis period (2007–2010).0068 (0.0036 0.188*** 0.013) (0.41 0. Guidara et al. (2) 1988 to 1997.00004) CR3 7.0007) (0. ** 5% Significance level.0185 1.0228 0.704) (13. * 10% Significance level.411) (0.0128* 0.25 0.227 20.0258 * 0.0003) (0.003) (0.215 (0. *** 1% Significance level. Model 1 (columns 1–3) includes the whole sample period 1982–2010.0304*** 0.683) (13.0141) DREG2 0.0183) (0. Capital buffers (BUFL) are calculated as the difference between shareholders’ equity-to-assets ratios and the inverse of the regulatory ceiling on an unweighted leverage ratio.0006) *** *** BUFLt1 0.0006) (0.0084) (0. The estimations are performed using two-step GMM regressions (2SGMM).510) DBUFL  DREG2 0.211 0. ulatory regimes: (1) the period before the Office of the Superinten- dent of Financial Institutions (OSFI) adopted Basel I guidelines.954 (0.32 (0.573) (0.0010) (0. .0010*** 0.524) (0.0051) DREG1 0. A.780) (0.0001*** (0. Values in parentheses are robust standard deviations. risk and performance. Market data was extracted from daily data and converted into quarterly units.0004) (0.308 0. we don’t need to adjust it by its mean. Financial data are quarterly observations from statements covering 1982–2010.009) TRISKt1 0.0038 0.667) (13. regarding the output gap.0007) (0.0212 0.75 (0.596) DBUFL  DREG1 0.70* (0. First.0345*** 0.08 0. in several respects.528 0.0033) (0. which introduced mar- capital buffers (the difference between the banks’ capital levels ket risk as a distinct risk category.0007 0.0009 0.0995* 0.814) DRET 0.0299 0.0027 0.00004) (0.0173 (0. Second. Conclusion from 1982 to 1987.09E05 (0.683) (13.81 0.0012*** 0. the banks’ risk and performance throughout business cycles and fect significantly our conclusions drawn from the results in Table 4.092 0.016) (0. First.0202 0.0017* 0.363 22.032) (0.0188** (0.0569 0. 6.687) (13.250) DTRISK 0.0009 0.709) (13.0091) BUFR 0. Second.0034 0.0010** 0.035) (0.0093) (0.0007) (0.008) TERM 0. and (3) 1998 to 2010.485 (0.0003 0.011) (0.014) (0.031) (0.0163** (0.199*** (0.0020** (0.007) CV 0.051) (1.575) (0.0009*** 0.002) BUFL CREDIT 0. This sample period allows buffer.001) VTSX 0.0073 (0. Model 2 (columns 4–6) only covers 1982–2006.055) Observations 565 565 565 510 510 510 R-squared 0.0004) (0.0144) (0.0517 (0.580) DREG2  OUTGAP 0.143 0. Model 1 (Period 1982–2010) Model 2 (Period 1982–2006) (1) (2) (3) (4) (5) (6) Variables DBUFL DTRISK DRET DBUFL DTRISK DRET OUTGAP 0. Regression and minimum capital requirements) and analyzes its impact on results presented in Table 9 show that these adjustments do not af.630) (0. dummy variable.088 Number of banks 6 6 6 6 6 6 This table presents regression results of systems of simultaneous equations of changes in bank capital buffers. Our work departs from the literature on capital buffers the cyclical component of real GDP obtained by applying the Hod.0031* 0.53 15.0003) (0.0001*** 0. thus the average us to account for at least three business cycles and three major reg- DBUFL is very close to zero.0066 0.012) (0.008* (0.0185) DREG1  OUTGAP 0. since we use regimes.0254* (0.053) RETt1 0.005) (0.0003) (0.673 17.052) (0.326*** 0. in response to Canadian regulatory changes during various Basel We explain as follows.45 0.527) DBUFL 0.

0333) (0. * 10% Significance level.009) (0.0196 (0.0026 (0.524) DBUFL 0. By and large.0003 0.031) (0. Guidara et al.006) (0.507** 0.0295 0.211 0.0014 0.43 (0.0060* 0.180*** 0.0003 0.0043 0.0270*** 0.136* 0.0012 0.266 2.051) (1. performance that variations of banks’ capital buffer impact banks’ exposure to and regulatory changes in the Canadian context constitutes an risks and return on equity.57E05) (3.313) (0.0850 1.593*** 3.0152 0.3386 A.129) (0.016) (0.0176) DREG1 ðOUTGAP  OUTGAPÞ 0.031) (0. their capital buffer? (4) How do induced changes in the capital buf- ital buffers with respect to business cycles and regulatory fer affect the performance of Canadian banks? changes—a question of paramount importance in the aftermath We find that Canadian banks are well capitalized.0076) (0. Model 1 is shown in columns 1–3.0008*** 0. In this table.0001) (0.418) DRET 0.0018*** 0.105 0.0033 (0.0074 0.0031* (0.138 0. ship between capital buffers and risk. rigorous and strict implementation of tion? (3) How sensitive are Canadian banks risk to changes in both risk-based and non-risk-based capital requirements can help .0453) RETt1 0.018) (0.040) (1.0003) (0.091 Number of id 6 6 6 6 6 6 6 6 6 This table presents regression results of systems of simultaneous equations of changes in bank capital buffers.004) (0.0156) (0. which helps of the subprime credit crisis—on the resilient Canadian banking explain why they weathered the recent financial crisis so well.0697) (0.64E05 1.0087) (0.0006) (0. First.0002) (0.0007) (0.003) (0. Canadian experience.001) VTSX 0.459) (0. we study bank capital buffer.011) (0.150*** 0.040** 0.005) (0.139 0. Financial data are quarterly observations from statements covering 1982–2010.005) (0.012) (0. risk and performance We document that bank capital buffers exhibit a positive co-move- simultaneously using a two-step generalized method of moments ment with business cycles. for changes in regulatory regimes.0009) (0.565 16.242 0.237) (0.515) (0.09 (0.0129* (0.0569 (0.606 0.0203 0.160 0.0068* 0. The values in parentheses are robust standard deviations. sector.086 0. business cycles.0397*** 0.67) DREG2 ðDBUFL  DBUFLÞ 0.002** 0.669 (0.58E05) (4.0003) (0.680 0. The estimations are performed using two-step GMM regressions (2SGMM).386) (0.0002) (0.701) (0.180*** 0.199) (5.16 (0.521 (0.24E05) CR3 3.0549) (0.573) DREG1 ðDBUFL  DBUFLÞ 0.007) CV 6.321) (0.0075 0.0099) (0. In the interaction terms.0817** 0.188*** (0.0169 0.0567 0.0008 0.0018) CREDIT 0. / Journal of Banking & Finance 37 (2013) 3373–3387 Table 9 Estimation results with an alternative method for the interaction terms. Hence. the motive to hold an We address the following research questions: (1) Do Canadian excess capital buffer may be driven by market discipline. Model 1 Model 2 Model 3 (1) (2) (3) (4) (5) (6) (7) (8) (9) Variables DBUFL DTRISK DRET DBUFL DTRISK DRET DBUFL DTRISK DRET OUTGAP 0.0093 0.0345*** 0.167) (0.0012*** 0.0004) (0.575) DREG2 ðOUTGAP  OUTGAPÞ 0.0114 0.0503 0.0093) TRISKt1 0.589 0.718 18. we use total equity risk (TRISK) as the risk measure and equity returns (RET) to measure performance.0103 0.015) (0. *** 1% Significance level.0366*** 0. Comprehensively addressing the relation.014) (0. We can then draw two main policy implications on the basis of dian banks’ capital buffers sensitive to changes in capital regula. We also find no strong evidence ship between capital buffers. there is no strong relation- important contribution to the literature. risk and performance.004) (0.0595*** (0. we adjust the variables OUTGAP and DBUFL by their means. model 2 in columns 4–6 and model 3 in columns 7–9.0013*** 0.001) (0.0134) DREG2 0. banks’ capital buffers run counter to business cycles? (2) Are Cana.0074) (0.749) (14.0001*** (4.467 0.0005*** 0.50E05 7.007) (0.0003) (0.0002) (0.007) (0.0012*** 0.0026 0.752) (14.0017* (0.305) (0.052) Observations 565 565 565 565 565 565 565 565 565 R-squared 0.45) DTRISK 0.012) (0.305 (0.0038 0.0097 * 0.002) (0.143 0.0013 0. our study is original in studying the cyclical behavior of bank cap.054) (0.91E05 8.003) (0.011) (0.012) (0.227 (0.0021*** 0.429 15.0010 0.0006) *** *** *** BUFLt1 0.0009 0. Market data was extracted from daily data and converted into quarterly units.005) DREG1 0. Third.105 0.201 0.0014*** 0.0008 0.214) (5. The capital buffer (BUFL) is calculated as the difference between the shareholders’ equity-to-assets ratio and the inverse of the regulatory ceiling on an unweighted leverage ratio.61) SIZE 0.0155** 0.009) (0. This result holds even when we control (2SGMM) framework.0019) (0.048) (1.233) (3.268) (0.326*** (0.0002) (0.243 0.0084) BUFR 0. ** 5% Significance level.007) TERM 0.0040** 0.727) (14.16E05** 0. All other variables are defined in Table 1.246) (3.0041 0.0008*** 0.005) (0.0594 0. risk.086 0.977 0.0007) (0.62E05 (0.0006) (0.177 0.0243* (0.114 0.056) (0.0416*** 0.

. 2008.. A.. St-Amant. conference of the Canadian Economics Association in Quebec City. P. Scale and scale effects in market-based accounting Etienne Bordeleau.... 2012. A. The impact of consolidation and safety-net support European Financial Management 13 (1).. Gregory.. Qi. Does stock option-based executive compensation induce risk-taking? An analysis of the banking industry. We thank Sandra Betton. P. E. Rome. Denver. 2008. 1–35. Bank of Canada Conference Proceedings. Strong boards. T. Canadian Journal of Economics 40. Crawford.M. Competition and contestability in Canada’s financial hua School of Management at the Peking University. 1989. Journal of Financial Stability 7 (2). Asset pricing. 1581–1603. 763–787.A. J. Journal of Banking and Finance 21.P. Beijing.. S. C. Kyriaki Kosmidou. G. Kwan. Haaf. 1999. 1998. the finance seminar at Jokipii. American Economic Integration: The Role of Risk-sharing Mechanisms. Journal of Banking and Finance 25. 2009. Nilanjan Basu. Levine. Pathan. Lai. Journal of Finance 41. Lindquist.. Leonard Nakamura. 2012. Efficiency and economies of scale of large Canadian banks. L. R.D. model. F... The relationship between risk and capital in provinces: 1961-2006. 2004. the 44th annual 247–273.. Économiques in Lac Beauport. Journal We thank an anonymous referee and the Editor Ike Mathur for of Banking and Finance 30. A. Working Paper 09/152. 49–70.A. 225–244. Discussion Paper growth because it is during these periods that banks can accumu. Jin-Chuan Flannery... Lessons for banking reform: a Canadian Soumaré would like to thank Professors Li Liu and Longkai Zhao perspective... Beck. Kretzschmar. 439–457. Journal of Money. USA. USA. simultaneous equations approach. McGraw-Hill. Harjes.. 1998. Blum... Cornett. A.. 2006. Bank of Shaffer. Dahl. Y.. Management Association. Capital requirements and bank behaviour: empirical evidence for References Switzerland. Altunbas. and Song. Jokipii. This paper International Money and Finance 23.. Anna Chernobai.. Remarks to the RBC of Singapore. 2001. 2003. Greg.C.. 391–429.... Graham. Pricing risk-adjusted deposit insurance: an option based Allen. R. A. 493–513.H. Journal of Banking and Finance 23.O. and Finance 33. C. Canada. Singapore.P. and seminar participants deposit insurance in Canada. Journal of Banking and Finance 30..W. Balli. Milne. concentration and their relationship: an Stolz. A... David Newton. Keshab Giammarino. Jokipii.. Claessens. Regulatory Constraints on Bank requirements should occur during periods of strong economic Leverage: Issues and Lessons from the Canadian Experience. S.S. 1990. Yaxuan 1990s? Review of Finance 12. 2004. 1997. Rangan. Eisenbeis. Ontario. S. and the Conference on Coun. 2003. American ties. Rajan. the Institut de Finance Mathématique de Montréal. A. R. system: empirical results.. research.. E. the RUC-EFMA Symposium on Asian Financial Manage. Office of the Superintendent of Financial Institutions Canada (OSFI).. Journal China. Risk-based capital. The cyclical behaviour of European bank capital buffers.. J. Huizinga. International Monetary Fund. tercyclical Capital Requirements of the Federal Reserve Bank of Journal of Financial Services Research 12.... T.. 2011. Wedow. Bank.R.. Paulin.. 1719–1728. 1993. 109–127. 2008. Jean Roy. V... . 25–56. crises: first results. tiative of the Risk Management Institute of the National University Dickson. Jean Louis. Capital and Procyclicality in a Turbulent Market. Rocco Huang. Zechner.J. 2009. Verma. E. A. rent Basel risk-based capital charges. R. Word Conrad Leblanc. on Canadian. provide more flexibility for banks to weather downturns.. Why are Canadian Banks more Resilient.. 323–340. M. Banks’ regulatory capital buffer and the business cycle: empirical analysis of the banking industry. Jacques. R.A. M. K. Canada Working Paper No. Huang. Bank concentration.. risk and efficiency in European banking. Anand Srinivasan. What caused the bank capital build-up of the ory Lipny. and bank capital: a ment organized by the Renmin University of China. 115–143.. 789–805. A. The New Basel Capital Accord and the Cyclical Behaviour of Bank Capital. Working Paper. conversely. 1999. Journal of Banking and Finance 34. Canadian Journal of Economics 22. and revenue diversification. 2009. the Multinational Finance Society.. 98–110.E. Competition. K. A. 1999-17. A two factor ARCH model for deposit-institution stock returns. Financial Institutions Management. A.. B. 2013. Calmès. Working Paper No. Channels of risk-sharing among Canadian Shrieves. A. Capital Markets Canadian Bank CEO Conference. C..M.. 1989. a reduc.. Nathan. the monthly workshop of the Risk Management Institute. 1440–1451. R. 2009... M. Risk Assessment and Employment and Canada Research Chair in Keeley.. Schwartz. Paulin. L. Saunders.. Swiss Journal of Economics and Statistics.. F. E.. Economic Review 5. McNeil. A. H. mes. A. T. Bank capital buffer and portfolio risk: the influence of business cycle Empirical Economics. Italy. Neave. Cleveland. competition. Journal of Business Finance and Accounting 30 (1-2). T. 49–61. 1994. R. Théoret. Guidara et al. Canadian Journal of Economics 22. P. Regulatory changes and financial structure: the case of Canada. Calmès.. S. Credit and Banking 25. M. 761–772. 2004. B. 2004-30. the Guanghua School of Management at Peking University.. Kirchner.. Swiss Society of Economics and tion in capital requirements would be desirable since it may Statistics 140 (1). S. C. Lai. S. Alexandre Jeanneret. P. B. Bond. China. risk.. Concordia University. Canada’s Exchange Rate Regime and North 537–571. M. Bikker. Balli.E. 165–178. 915–945. 1699–1707. 2007. late more capital. 2010.. White. 576–594. 2004. A. Christian Cal. C. R. time-varying risk premia Duan.. 2191–2214. The views expressed in this paper are those of the authors Economic Review 88. Dipinder Randhawa. for their valuable comments. International Review of Economics and Finance 12.. Zingales. A. S.. McGraw. University of Singapore.. Market valuation of bank assets and Shrestha. K. 533– the annual meetings of the Midwest Finance Association. 2006. Toronto. 1997. Bank risk. Steiner. Canada. Quebec. CEO power and bank risk-taking. J. R.. USA. How does Foreign Entry Affect their valuable comments. Second. the FSA ULaval Research Office. US and UK banks: 1893-1992. Bank of Canada. Journal of Banking and Finance 32 (8). Flannery. P..J.. Journal of Economics and Business 49. relationships between capital. G.G. for their financial support. Y. 559–589. China. Astrid Juliane Salzmann. S. 105–127.A. the Credit Risk Ini.. Beijing.. Journal of Money Credit and Banking 26. 2007. We are grateful to the Fonds d’Assurance the Domestic Banking Market. E. Min-Teh Yu. Gueyie.. Illing. 2009. Z. Ratnovski. 1998.. Bank of Canada. 315–335. Cedric Okou.. H. Nonlinearity of Bank Capital and Charter Values. Journal of Banking tality..M. January 8. of Financial Stability 7 (3). D. Carbo. increases in capital Bordeleau. M. commercial banks. Demirgiic-Kunt. Journal of Banking and Finance 34. Integrated models of capital adequacy: why banks are undercapitalized. Deposit insurance. Chicago. Blundell. the Fonds Québécois de la Recherche sur la Société et la Culture D’Souza. Bank moral hazard and the introduction of official at the 50th annual meeting of the Société Canadienne de Sciences deposit insurance in Canada.. 1183–1200. American Risk Management Joint Seminar at HEC Montreal. 2010.. 1918. R. Djoudad. Hameed. S. Molyneux. 1986. Journal of Banking and Finance 37 (3). Initial conditions and moment restrictions in dynamic panel data models. and not necessarily those of their institutions.. T.J. Liu. and market power in banking... the Financial Swiss National Bank. Ryerson. Interaction effects in econometrics. Forthcoming in: Shim. J. Boubacar Camara. Examining the third Canadian ed. A test of competition on Canadian banking. Evren Damar. Northcott. Milne. during recessionary times. 2009-15. Beijing. 42–53. 2009. Rime. 547. Wilson. Bank capital buffer and risk adjustment decisions. Acknowledgments Chen. 2011. Empirical Economics 43 (2).. 2002. evidence for Germany. 117–131. Journal of Banking and Finance 32 (8). 1340–1350... the finance sem. National 2838–2850.. Basher. capitalization and operating efficiency. Ronn. inar at the John Molson School of Business. Saunders. 871–895. Antia. 1992. Whyte. Sorensen. was in part finalized while Issouf Soumaré was visiting the Guang. R. M. the Inter-University Economic Centre on Economic Policies. Journal of Banking and Finance 16. C. J. Journal of Banking and Finance 26. Christian Dorion. 2006. Does diversification improve bank efficiency? In the and the Social Sciences and Humanities Research Council of Canada evolving financial system and public policy. Gardener. Journal of Econometrics 87... / Journal of Banking & Finance 37 (2013) 3373–3387 3387 mitigate the well-documented procyclicality associated with cur. All errors and omissions are the authors’ sole responsibili. Policy Research Working Paper No. Central Banking Journal 19 (4). portfolio risk. 1997. Why Basel II may need a leverage ratio restriction. J. Martin Boyer.. Banks’ buffer capital: how important is risk? Journal of Cleveland. The impact of off-balance-sheet activities on banks returns: an application of the ARCH-M to Canadian data. Financial dependence and growth. Easton..L.R. Nigro. G. and interest rate risk. J. and other members of the Department of Finance for their hospi. Demirguc-Kunt.