Financial Expertise of the Board, Risk Taking, and Performance: Evidence from Bank Holding Companies* Bernadette A. Minton The Ohio State University

Financial Expertise of the Board, Risk Taking, and Performance: Evidence from Bank Holding Companies* Bernadette A. Minton The Ohio State University Jérôme P. Taillard Boston College and Rohan Williamson Georgetown University July 2012 Abstract Financial expertise among independent directors of U.S. banks is positively associated with balance-sheet and market-based measures of risk in the run-up to the 2007-2008 financial crisis. While financial expertise is weakly associated with better performance before the crisis, it is strongly related to lower performance during the crisis. Overall, the results are consistent with independent directors with financial expertise supporting increased risk-taking prior to the crisis. Despite being consistent with shareholder value maximization ex ante, these actions became detrimental during the crisis. These results are not driven by powerful CEOs who select independent experts to rubber stamp strategies that satisfy their risk appetite. Key Words: Governance, Risk Taking, Board Composition, Financial Expertise, Risk Management Committee JEL Classification: G20, G21, G24, G32 *Earlier versions of this paper were titled, Do Independence and Financial Expertise of Board Matter for Risk Taking and Performance. We thank an anonymous referee, Viral Acharya, Laura Fields, Laurent Frésard, Kai Li, Brian Peters, Jeff Pontiff, Jun Quian, Philip Strahan, David Scharfstein, René Stulz, and seminar participants at the NBER Fall 2010 Corporate Finance conference, the Annual Corporate Governance Conference at the Weinberg Center for Corporate Governance, the 8 th International Paris EUROFIDAI-AFFI Finance Meeting, the Conference on Governance and Risk Management in the Financial Services Industry at the FRBNY, the Western Economic Association Meetings, Boston College, Georgetown University and Ohio State University for helpful comments and suggestions. We also thank Kelvin Moses, José Pérez, Robert Prilmeier, Felipe Restrepo, and Mai Zhang for excellent research assistance. Minton acknowledges financial support from the Dice Center for Research in Financial Economics. Williamson acknowledges financial support from the Stallkamp Faculty Research Fellowship and the Center for Financial Markets and Policy at Georgetown University. Authors email addresses: minton_15@fisher.osu.edu, taillard@bc.edu, williarg@georgetown.edu 1

Introduction The failure of various internal governance mechanisms has often been cited among the key contributing factors to the 2007-2008 financial crisis (see for example Bebchuk and Spamann (2010), Guerrera and Thal-Larsen (2008), Hashagen et al. (2009), and Strebel (2009)). In particular, calls for reform of the financial sector argue that the lack of financial expertise of board members played a major role in the crisis (e.g. Kirkpatrick (2009) and Walker (2009)). Financial expertise among independent directors is low among U.S. financial institutions. A quarter of all publicly-traded U.S. commercial bank holding companies with over one billion dollars in assets do not have a single financial expert among their independent directors. 1 As such, it is commonly believed that the presence of more independent financial experts on the board would have limited the excessive risks taken by banks management and mitigated their fall during the financial crisis. The goal of this paper is to test this assertion by analyzing how the banks measures of risk and performance relate to the presence of independent financial experts on their board. Financial experts among a board s independent directors arguably have lower costs in acquiring information about the complexity and associated risks of certain financial transactions and hence are better able to efficiently monitor senior management (Harris and Raviv (2008)). As such, a more financially knowledgeable board can recognize risks that will not pay off or which are unsound for the financial stability of the bank and advise senior managers to avoid such risks. Alternatively, these financial experts can identify risks that are more beneficial to shareholders in normal times and encourage management to take on those risks. Additionally, recognizing the explicit and implicit government guarantees given to banks, financial experts 1 For example, at year-end 2006, JP Morgan Chase had 12 independent directors of which only two had financial expertise. See Appendix A for details on JP Morgan Chase. 2

among the independent directors might encourage senior management to pursue risk taking activities to increase the residual claims of the bank s shareholders (Merton (1977), Marcus (1984), Keeley (1990), Acharya et al. (2010)). Thus, it is an empirical question as to whether the presence of financial experts on the board is associated with more or less risk taking and whether the risks taken create value for shareholders. In this paper, we examine these questions by analyzing a large sample of U.S. commercial bank holding companies both prior to and during the 2007-2008 financial crisis. After controlling for various bank, board, and risk management characteristics, we find that the fraction of independent financial experts is positively related to several measures of risk in the period leading to the 2007-2008 financial crisis. Our results are stronger for the subset of large banks, defined as having total assets greater than those of the median bank. 2 Specifically, an increase in the percent of financial experts among independent directors is positively related to the total risk of the bank s equity and its real estate loan exposure and negatively related to its Tier-1 and total risk-weighted capital ratios. For example, a one-standard deviation increase in the our financial expertise measure is associated with a 34-basis point decline in a bank s Tier-1 risk-weighted capital ratio; an economically significant drop compared to the average Tier-1 risk-weighted capital ratio of 10.4% among large banks. Moreover, we observe that prior to the crisis, between 2003 and 2006, banks with more financial expertise slightly outperformed banks with less expertise. These results are largely consistent with the view that financial expertise is associated with more risk taking that benefits shareholders in normal times. Yet, when we analyze how financial expertise relates to banks performance during the 2007-2008 crisis, we document a strong negative relation that holds especially for large banks. 2 Our sample consists of bank holding companies with total assets greater or equal to $1Bn. For the period between 2003 and 2008, the median value for total assets stands above $3Bn dollars. 3

A one standard deviation increase in financial expertise levels is associated with a five percent drop in stock value during the crisis; an economically meaningful impact when compared to a median stock performance of about minus 40% between January 2007 and December 2008. For the subset of large banks, the negative relation remains statistically and economically significant even after augmenting the set of controls to include exposure to real estate loans, deposit financing, and mortgage-backed securities. Taken together, these findings indicate that financial expertise is associated with more risk taking that potentially benefits investors before the financial crisis but turn out to be detrimental during the crisis. These results are consistent with Beltratti and Stulz (2010) who study an international panel of large banks and find that pro-shareholder boards are associated with higher (lower) performance prior to (during) the crisis, potentially reflecting decisions that were thought to maximize shareholder value but did not perform as expected when the crisis hit. Endogeneity concerns always exist in studies relating corporate governance aspects to firm performance (e.g. Hermalin and Weisbach (1998), Adams, Hermalin and Weisbach (2010)). In our setting, directors with financial expertise are not randomly allocated across banks. Empirically, we specify lagged independent variables similar to the main specifications of Coles, Daniel and Naveen (2008) and Linck, Netter and Yang (2008) and separately use propensity score matching techniques, as in Drucker and Puri (2005). The results of our propensity score matching analyses corroborate our regression-based results. We also test a potential reverse causality interpretation of our results by examining whether the presence of independent financial experts on boards is positively related to the presence of a powerful CEO. The rationale behind the focus on powerful CEOs is that if boards are ineffective, it is because the top management, and in particular the CEO, have control of the 4

firm (Hermalin and Weisbach (1998)). In our setting, it might be that powerful CEOs select independent financial experts to rubber stamp strategies that satisfy their risk appetite. For this test, we use four proxies of CEO power and interact these proxies with the financial expertise variable to test whether the negative relation found between performance during the crisis and levels of financial expertise is driven by banks with powerful CEOs in place. The interaction terms are not significant in any of the specifications, thus providing evidence that rubber stamping by independent financial experts is not driving our main results. Furthermore, we verify whether the Troubled Asset Relief Program (TARP) in which the government created a fund that protected against systemic risk had an impact on our results. Though the likelihood of receiving TARP is associated with certain bank characteristics and board independence, it is not associated with financial expertise of independent directors. The results are robust to many alternative specifications. First, our definition of financial expertise follows Güner, Malmendier and Tate (2008) but we also run our main tests using alternate definitions of financial expertise and the results hold. Second, we show that all of the main results hold when independence is defined more restrictively, by taking into account other potential social and business ties between a director and the CEO of the bank (see Fracassi and Tate (2010)). Third, the results hold when we exclude the five largest commercial banks, and hence are not driven by unique features of the mega-banks in our sample. Lastly, the results hold when the sample includes Savings & Loans associations (S&Ls) and investment banks. Our research extends the literature examining the role of internal governance mechanisms going into the financial crisis 3 by looking at a specific aspect of board governance financial 3 Several recent papers examine the role of internal governance mechanisms going into the financial crisis (e.g. Adams (2009), Erkens, Hung, and Matos (2009), and Ferreira, Kirchmaier, and Metzger (2010)). For instance, a fair amount of attention has been given to the incentives of bank management in the run-up to the crisis and during the 5

expertise of independent directors that is relevant to the current debate on how to fix financial firms and regulate the industry going forward. 4 In stable times, the presence of financial experts among the independent directors is associated with higher risk taking and slightly above-average performance. Since financial expertise on the board is related to more risk taking, it is not surprising that banks with more independent financial experts underperform when the crisis hits. These results are still consistent with the board acting to maximize shareholder value ex ante. However, our results do challenge the popular view that more financial expertise on the boards of banks would unambiguously lower their risk profile since this lower risk level may not have been in the best interest of shareholders prior to the crisis. The remainder of the paper proceeds as follows. Section I describes the sample and board characteristics. Section II discusses results related to risk taking and board composition in the run-up to the 2007-2008 crisis. Performance prior to and during the crisis and how it relates to board composition are examined in Section III. Section IV discusses several robustness tests. Section V concludes. I. Sample description and board characteristics I.A Sample crisis (see Fahlenbrach and Stulz (2011), Bebchuk, Cohen, and Spamann (2010), and Cornett, McNutt and Tehranian (2010)). 4 In a recent study, Fernandes and Fich (2009) examine how financial expertise relates to bank performance during the crisis. Their definition of expertise is quite different from ours as they define financial expertise of the board as the average number of years of banking experience among board directors. Overall, their results are different from ours. We conjecture that this is due to several factors including the different set of banks and controls they use, the different definition of financial expertise and their different empirical strategy; in particular, they focus on an event study around Lehman s bankruptcy. 6

The original sample of financial institutions includes all U.S. banks along with specialty and other finance firms on the BoardEx database, a business intelligence database that includes profiles of over 350,000 business leaders across 19 countries, during the 2000 to 2008 period. Given the smaller and potentially biased coverage prior to 2003, we focus only on the 2003 to 2008 period. We eliminate small local institutions by focusing only on those with more than one billion dollars in assets. All insurance companies, REITs and investment companies are also eliminated by keeping only those firms in the Standard Industrial Classification codes of 602, 603, 6211 and 6282. The remaining sample includes commercial bank holding companies, S&Ls and investment banks. Commercial bank holding companies comprise 75% of the sample, S&Ls about 23% and investment banks 2%. One concern related to the empirical analysis is the potential heterogeneity among different types of financial institutions. The impact of the financial crisis arguably might differ across commercial banks, S&Ls, and investment banks. To make sure that there is consistency across the sample, we focus our main analysis on the subset of commercial bank holding companies (hereafter referred to as banks). We discuss the impact on investment banks and S&Ls relative to banks in the robustness section. The sample of banks varies from 182 banks in 2003 to 119 in 2008, with a maximum of 206 banks in 2005. We also require that banks have data on Compustat and CRSP. 5 Lastly, much of the discussion surrounding the crisis and how the government would respond to each distressed financial institution has revolved around the idea that large institutions are different from small institutions. In particular, if implicit government backing is stronger for large banks, it might affect ex ante how board members view the optimal risk taking level. 5 Depending on the specifications of our tests, the sample size is sometimes reduced further due to data availability issues on some balance sheet items, stock market data or call report data. 7

Furthermore, a bank s size can affect its risk profile in various ways. Larger banks tend to be better diversified and hence less vulnerable to shocks than smaller banks. Conversely, larger banks might also engage in more risky activities than do smaller banks, such as more off-balance sheet operations. Thus, in the empirical analyses we run each set of tests on both the full sample of banks and the subsample of large banks, defined annually as banks with total assets exceeding those of the median bank. I.B Financial expertise Information on the financial expertise of independent directors is collected from annual bank proxy statements and BoardEx. Similar to Güner, Malmendier and Tate (2008), we classify an independent director as a financial expert if he or she (i) has worked within a banking institution, (ii) currently works at a non-bank financial institution, (iii) has a finance-related role within a non-financial firm (e.g. CFO, accountant, treasurer, or VP finance) or (iv) academic institution (e.g. professor in finance, accounting, economics or business), (v) is a professional investor (e.g. hedge fund, private equity). It is common for a director to sit on the board of both the parent bank holding company and one of its subsidiaries. 6 However, bank regulations do not allow a management official of one depositary institution to sit on another bank s board at the same time if they are not affiliated. Accordingly, we do not classify a current executive working at a subsidiary of the bank to be an independent financial expert at the bank holding company level. 7 Our financial 6 In a majority of cases, it is related to acquisitions where an executive of the acquired banks steps down and joins the board of the merged entity (see Adams and Mehran (2003, 2008, and 2012). 7 We thank the referee for making this point and helping us clarify our previous definition of financial expertise. Although exemptions can be made, Federal Reserve regulation L forbids interlocks of management officials for depositary institutions and their holding companies. 8

expertise variable is measured as the fraction of reported independent directors that are classified as financial experts. 8 Figure 1 summarizes the average composition of financial expertise over the sample period. As Figure 1 shows, more than 30% of independent directors with finance expertise hold finance related positions at non-financial firms and more than half of the financial experts are current or former executives at financial firms, with 43% holding positions at non-bank financial institutions. While regulations do not allow bankers to work and be a board member in two different banks at the same time, several banks place former bankers on their board. In all, more than 10% of independent financial experts have past executive experience in banking. [Insert Figure 1 here] Table I reports data on financial expertise and other board characteristics. Column one in Panel A of Table I summarizes the average level of financial expertise across the sample. The fraction of financial experts among independent directors increases from 20% in 2003 to 26% in 2008. [Insert Table I here] Figure 2 shows the distribution of independent financial experts among banks at year-end 2006. It is interesting to note that the distribution is relatively spread out. Furthermore, the figure shows that almost 25% of banks have no financial experts among their independent directors. Given this high proportion of zero financial expertise among independent directors, 8 See section V for results with alternate definitions of financial expertise. 9

we also run all of our tests defining financial expertise as an indicator variable that takes value one if at least one independent director has financial expertise and zero otherwise. All the results using this variable are qualitatively similar to those reported and are therefore not shown. The propensity score analyses also use this definition to categorize banks in the treatment versus the control group. [Insert Figure 2 here] I.C Other board characteristics We control for three other characteristics of banks boards: board size, board independence, and the presence of a risk management function. 9 Data on board size and board independence are taken from BoardEx which reports the disclosure made by firms along those dimensions. Board size is defined as the number of directors on the board. Decision making costs are often assumed to be lower in smaller boards than in larger boards. As such, smaller boards might be more effective monitors. Board independence is the ratio of the number of directors reported as independent in a bank s proxy statement relative to board size. Independence is based on the definition set by the NYSE and NASDAQ guidelines. As shown in Table I, the average board size decreased slightly from about 13.3 board members in 2003 to about 12.4 members in 2008. Levels of board independence are high among our sample of banks. In 2003, 73% of the board is independent. By 2008, independent directors make up over 78% of the board. Ellul and Yerramilli (2010) and Aebi, Sabato, and Schmid (2011) show that risk management governance can matter for risk taking. For almost all commercial banks, the risk 9 Busy boards are related to poor corporate governance among non-financial firms (Fich and Shivdasani (2006)). We do not control for this characteristic as it is a rare feature among our sample of commercial banks. 10

management function is performed either by the Audit Committee or the Asset and Liability Management Committee of the board of directors. Thus, there would be no variation in indicator variables denoting these committees across banks. However, some banks disclose the presence of a Chief Risk Officer (CRO) or Risk Management Committee that oversees risk monitoring and reports to the Board of Directors and separate from the Audit or Asset and Liability Management Committees. To detect the presence of these two risk monitoring functions, we perform multiple key word searches of all 10-Ks and DEF 14-A (proxy) statements during the sample period. We then verify manually each flagged financial statement and hand-collect all the disclosed information regarding both the executive positions and committees related to risk management. By relying on voluntarily disclosed information by firms, we are probably underestimating the true number of firms having these risk functions in place. The risk management structures (CRO, and risk committee, discussed below) are also positively related to bank size (untabulated). This fact is not surprising given the increased complexity of operations at larger banks. Panel B of Table I presents a correlation matrix for the variables measuring board size, board independence, finance expertise, and selected bank characteristics. There are several features of the data worth highlighting. First, there is a positive correlation between board size and bank size, which is consistent with the results presented in Adams and Mehran (2003, 2008). Second, there is no correlation between our measure of finance expertise of the board and other board characteristics. Hence financial expertise among independent directors reflects a dimension of the board that is not captured by other board variables. Lastly, there is a slight negative correlation between bank size and the fraction of financial experts among independent directors. 11

II. Board composition and firm risk In this section, we examine the associations between board characteristics and two groups of risk measures in the run-up to the 2007-2008 financial crisis: a market-based measure of risk (section II.A) and two balance sheet-based measures of risk (section II.B). To examine whether large banks behave differently from small banks, we run each test on the subset of large banks separately. We also investigate the impact of sample biases by using a propensity score matching technique to address possible endogeneity concerns (section II.C). II.A Board composition and market-based risk Table II presents the OLS regression coefficient estimates for regressions examining the relation between total risk and bank and board characteristics. Total risk is our market-based measure of risk and is computed annually as the standard deviation of daily stock returns over the calendar year. The dependent variable in the OLS regressions is the average total risk over the 2004 to 2006 period. All explanatory variables measuring bank and board characteristics including the risk monitoring function, board size, board independence, and financial expertise are measured at year-end 2003. [Insert Table II here] Results for all banks are reported in the first two columns and for the subset of large banks in the last two columns. 10 For each sample, we estimate two regressions: one excluding 10 To be consistent throughout the study, the above-median cutoff for large banks is the same as in the other tables. However, there is a larger fraction of the smaller banks that do not have all the data for this specification; this explains the low sample size reported in the first two columns of the table. 12

the fraction of financial experts among independent directors (Regression (1)) and one which includes this financial expertise variable (Regression (2)). As the table reports, larger banks and larger boards are associated with lower levels of total risk. For example, for the full sample of banks, a one member increase in board size is associated with a decrease in total risk of 15 basis points. This decrease has to be compared to the average daily total risk for the sample of firms is equal to 1.25%. The negative relation between risk and board size is similar to the results in Pathan (2009) for financial firms and Cheng (2008) for non-financial firms. Cheng (2008) argues that because of the compromises needed to reach decisions in larger boards, stock returns should be less variable for firms with larger boards. Table II shows that, after controlling for bank and other board characteristics, financial expertise among independent directors is positive and statistically significantly related to total risk (Regressions (2)). The coefficient estimates, however, are larger and more statistically significant for the subset of large banks than they are for the full sample of banks. For the subset of large banks, a one-standard deviation increase in the fraction of financial experts among a board s independent directors is associated with a seven-basis point increase in average daily volatility, corresponding to slightly more than a 5% increase in total risk relative to the sample mean. For the full sample of banks, a one-standard deviation increase in the fraction of financial experts among a board s independent directors is associated with less than a four-basis point increase in average daily volatility, corresponding to a 2.5% increase in total risk relative to the sample mean. The positive associations are consistent with financial experts acting in the interest of shareholders as shareholders benefit from more risk-taking given that their cost of capital does not reflect the riskiness of their assets (Merton (1977)). The positive associations are also 13

consistent with a more financially knowledgeable board having a better understanding of complex investments and potentially encouraging bank management to increase their risk taking. II.B Board composition and balance sheet-based measures of risk We analyze two balance sheet choices in this section: real estate related activity and bank leverage. 11 Data on the bank balance sheets are from FR Y-9C filings and are available from the Federal Reserve Commercial Bank Holding Company Database. We examine real estate activity because of its prominent role in the financial crisis. Bank leverage is examined through the lens of capital ratios. Given that equity capital serves as a buffer to weather financial crises, capital ratios are direct measures of the willingness of the bank to take on a riskier profile. Commercial banks have discretion in choosing levels of equity capital, as long as it is above the minimum regulatory levels. 12 The higher the capital cushion, the safer is the bank. However, from a shareholders perspective, the higher levels of capital dampen the potential returns on equity. The cross-sectional regression results for capital ratios and real estate exposure at the onset of the crisis are shown in Table III. As in Table II, we estimate the OLS regressions for the full sample of banks and the subset of large banks. [Insert Table III here] 11 See the robustness section for the discussion of results along other dimensions of risk taking. 12 The distribution of Tier-1 capital ratios is bell-shaped and does not exhibit any lower bound spike. 14

Panel A of Table III provides the OLS regression results for Tier-1 capital to the riskadjusted assets ratio. 13 Similar to the regression in Table II, we estimate two sets of regressions: one without the percentage of financial experts among independent directors (Regression (1)) and one with this variable (Regression (2)). Consistent with extant research, lagged net interest margin is negatively associated with equity capital across all regression specifications. Higher levels of profitability are associated with lower capital buffers. After controlling for bank and other board characteristics, financial expertise is negatively related to Tier-1 risk-weighted capital ratios in the subset of large banks. A onestandard deviation increase in the percent of financial experts among independent directors is associated with a 34-basis point decline in a bank s Tier-1 risk-weighted capital ratio; an economically significant drop compared to the average Tier-1 risk-weighted capital ratio of 10.4% among large banks. This negative association is consistent with the idea that banks with more financial experts had more leverage and hence were more exposed at the onset of the financial crisis. Lastly, as Panel A reports, the other board variables are unrelated to the banks capital ratios. Panel B of Table III provides the regression results for exposure to real estate, defined as the proportion of loans secured by real estate on the balance sheet. Surprisingly, financial expertise, as well as the other board characteristics, is not related to real estate exposures of banks across most regression specifications. II.C. Association between financial expertise and risk taking 13 Results explaining Tier-1 capital to total assets and total capital to risk-adjusted assets yield quantitatively similar results and are hence not shown. 15

As reported in Table I, about 25% of a board s independent directors are financial experts. The results in Table II and Table III show that, after controlling for bank and board characteristics, including the risk management function, these experts, on average, are associated with higher market- and balance-based measures of risk, especially for large banks. In this section, we address the interpretation of these results and in the following section we examine the performance consequences of these associations. One interpretation of our results is that independent financial experts, representing on average a quarter of the independent directors on the board, influence and/or sanction the decision making of senior bank executives such that the banks measures of risk are higher. Another interpretation, more skeptical and which recognizes regulators monitoring of banks activities and capital positions, is that independent directors with financial expertise are not randomly allocated across banks and many of the variables that are correlated with the appointment of these directors are also correlated with the banks risk profile. In the absence of a natural experiment or an econometrically valid instrument, we follow Drucker and Puri (2005), Aggarwal, Erel, Stulz, and Williamson (2009), and Fahlenbrach, Minton, and Pan (2011) and use propensity score matching techniques to address the potential endogeneity concerns. [Insert Table IV here] Panel A of Table IV presents the results of propensity score analysis for total risk. Panel B and Panel C present the results of the propensity score analysis for balance sheet risk measures. In the first stage, we calculate each observation s propensity score. The propensity score is equal to the probability that a firm with given characteristics has a financial expert 16

among its independent directors. To be consistent with our regression results in Tables II and III, the set of controls in the first stage are the same bank and board characteristics used as controls in the regression framework. In the second stage, each firm with a financial director among its independent directors (treatment group) is matched with a set of firms that do not have a financial director among its independent directors (control group) that have the closest propensity score to the treated firm. 14 The results in Panel A of Table IV show that the average total risk during the period 2004-2006 is greater for banks with at least one independent financial expert on their boards. The difference between the two groups is statistically significant for the full sample of all banks but not for the subset of large banks. Generally, the results indicate that banks with more independent financial experts had higher total risk prior to the crisis. Panel B of Table IV reports the results of our propensity score analysis of Tier-1 riskweighted capital ratios. Large banks with financial experts among their independent directors have lower Tier-1 risk-weighted capital ratios. In contrast, there is no difference between Tier-1 risk-weighted capital ratios between the treatment and control group of banks for the full sample. These results are consistent with those shown in Table III using a regression approach. The propensity score analysis in Panel C shows that large banks with financial experts among their independent directors have higher real estate exposure than those with no financial expertise among their independent directors. As the propensity matching approach allows for a better match among treated and control banks based on observable differences such as bank size, we can conclude that there is a significant link between the presence of independent financial 14 Results are shown for the radius methodology. Similar results are obtained when using the kernel, caliper or nearest neighbor match (5) approach. 17

experts on the board and higher risk-taking profile as captured by higher leverage and higher exposure to real estate among the subset of large banks. In this sub-section we have focused on potential endogeneity concerns of our results. Overall, we show that banks with more financial expertise were engaged in more risk-taking in the period leading up to the financial crisis which is consistent with results shown in the previous sub-sections. The risk level was higher for the total risk as well as for risk capital and real estate which is believed to be one of the main drivers in the financial crisis. III. Board characteristics and performance In the previous section, we show that banks with more financial expertise took on more risks in the period leading up to the financial crisis. In this section, we examine the consequences of these decisions by examining the associations between financial expertise of the board and performance during the 2007-2008 crisis period, as well as during normal times prior to the crisis. 15 III.A Performance during the 2007-2008 crisis period Table V presents the regression coefficients of stock performance during the crisis period on bank and board characteristics. Stock performance is measured as the nominal cumulative stock return from January 2007 to December 2008. 16 The set of firm and board characteristics used as controls in the regressions are measured at fiscal year-end 2006, except for banks past 15 We do not use accounting-based measures of performance such as write-downs on toxic assets as these numbers can be booked at different times for different banks, sometimes reversed and are generally subject to changes in accounting rules and government intervention. 16 We require at least six months of valid monthly stock return data in order to compute the cumulative returns. Therefore, the sample includes banks that delist between June 2007 and December 2008, as long as they were trading since January 2007. Results using abnormal returns without beta as a control yield quantitatively similar results. 18

cumulative stock returns measured during the calendar year of 2006 and beta, which is computed using a market model with daily returns over 2006. [Insert Table V here] Panel A of Table V gives the results for the full sample of banks. The first two columns contain a base set of regressors and the second two columns contain an augmented set of controls. The set of augmented controls includes the same variables as those used in the first two columns and adds several extra dimensions of risk exposures from Call Report (FR Y-9C) data: The proportion of total loans to total assets, the proportion of deposits to total assets, a measure of real estate exposure with real estate loans to total assets, the amount of non-deposit short-term financing, the level of off-balance sheet securitized assets (home equity, car loans, credit cards, etc.) and a measure of exposure to mortgage-backed securities with the total amount of mortgage-backed securities held on the balance sheet (including those held by the firm, available for sale and in the trading account) relative to total assets. These data are all winsorized at the 1% and 99% levels. As Panel A reports, few firm and board characteristics are significantly related to performance. Beta is positively and significantly related to stock performance over the crisis period in all specifications. The presence of a Chief Risk Officer (CRO) has a positive relationship with firm stock performance and is significant for three of the four specifications, while the percent of independent directors is significantly and negatively associated with stock performance during the crisis. These results are confirmed in Aebi, Sabato and Schmid (2011) who examine the impact of board level risk management on bank performance during the 2007-2008 crisis. 19

Holding firm and board characteristics constant, the percentage of independent financial experts and board size are not significantly related to stock performance during the crisis when using the base set of controls in the first two columns. We perform several robustness checks including different control variables such as the credit rating of financial institutions. For instance, if beta is taken out, then other firm characteristics, such as equity capital, become significant in explaining stock returns during the crisis. However, our results for financial expertise hold across all specifications. In columns three and four of Table V Panel A, we include the set of variables that reflect financing and other operating and financing choices made by the bank. The results show that independence is still negatively associated with performance, while financial expertise is not related to worse performance during the crisis. Not surprisingly, exposure to real estate is strongly and negatively related to performance during the crisis. The results for large banks are shown in Panel B of Table V. For the large bank subsample, beta is statistically significant for only two of the four specifications and all other base controls are not significant. The only firm characteristic that is significant throughout is the exposure to real estate as measured by real estate loans to total assets. The negative association between real estate loans and performance is not surprising given the central role of the real estate bubble in the financial crisis. In terms of board characteristics, similar to the results in Panel A, board independence is negative and significant for all specifications though less so for the base specification. In contrast to the full sample of banks, in all specifications in Panel B, financial expertise among independent board members among large banks is negatively related to stock performance during the crisis. Moreover, the associations are economically and statistically significant. A one 20

standard deviation increase in financial expertise levels at the onset of the crisis is associated with a five percent drop in the stock market performance during the crisis, relative to a mean (and median) stock performance of about -40% between January 2007 and December 2008. The negative association remains when we use the augmented set of controls in the regressions. 17 III.B Performance during the 2003-2006 pre-crisis period Table VI reports the analyses that relate bank and board characteristics to bank performance over the 2003 to 2006 pre-crisis period. The dependent variable is the cumulative stock return during the pre-crisis period. All independent variables are taken as of fiscal yearend 2003. [Insert Table VI here] Table VI, panel A shows the results for the full sample of banks. As in Table V, the first two columns contain a base set of regressors and the second two columns contain an augmented set of controls. In contrast to the crisis-period regressions in Table V, the coefficient estimates on the financial expertise variable are positive and statistically significant. A one-standard deviation increase in the percent of financial experts among the independent directors is associated with a 13% increase in stock returns, compared to a median stock appreciation of 70% over the pre-crisis period. Panel B reports the OLS regression results for subset of large banks. Similar to the results for the full sample of banks, the coefficient estimates on the financial expertise variable 17 To further confirm these results, we also run the same set of tests explaining changes in firm value as measured by percentage changes in Tobin s Q (Market to book value of assets) during the crisis period, from fiscal year-end 2006 to fiscal year-end 2008. This specification requires banks to be in the sample at fiscal year-end 2008, which reduces the dataset. The results are very similar to those from the stock performance regressions and hence not shown. 21

are positive. However, they are not statistically significant. Overall the pre-crisis results are consistent with banks with more financial expertise on the board being associated with better stock performance leading up to the financial crisis. III.C Risk taking, expertise and CEO power Given the cross-sectional nature of our tests, one possible competing explanation for our findings is that strong CEOs who were more prone to risk taking hired independent financial experts to rubber stamp their intentions. If that were true, the CEO s desire to take on more risks is driving both an increase in risk taking (and a consequent poor performance during the crisis) and higher levels of financial expertise among independent directors. This rubber stamping from independent directors could give the illusion of independent expert verification, while in reality the powerful CEO would control the board s decisions. We test this alternative explanation for our results in this section. To run these tests, we define four proxies to capture CEO power relative to their boards. The first measure (Table VII, Column 1) is an indicator variable that takes the value of one whenever the CEO has a dual role of CEO and Chairman of the board and zero otherwise. The dual role of a CEO has long been recognized as giving the CEO significant power over the board (e.g. Dahya et al. (2002)). The second measure is taken from Landier et al. (2007). They define a director to be independent of the CEO if his appointment to the board occurs prior to the appointment of the current CEO. A measure of board strength can be derived from the percentage of board members whose tenure on the board exceeds that of the current CEO. We rank all banks according to this measure and define a CEO strength indicator variable that takes the value of one if the bank falls below the median in terms of this 22

measure of board power (Table VII, Column 2). The other two measures are closely related to the one just described. Instead of computing the percentage of all directors with tenure greater than the current CEO, we compute the percentage of independent directors with greater tenure and again define firms with powerful CEOs with an indicator variable that takes the value of one for all banks below the median of the commercial bank sample (Table VII, Column 3). Lastly, we compute the percentage of independent financial experts whose tenure exceed that of the CEO and also define a CEO strength variable to take the value of one for all banks that fall below the median value according to that dimension (Table VII, Column 4). The correlation among these different measures of CEO strength is positive. Among the three measures based on relative seniority of board members to the CEO, correlations range between 0.80 and 0.44 and between these seniority-based measures and the dual role indicator variable, correlations vary between 0.23 and 0.16. [Insert Table VII here] The results in Table VII are qualitatively similar across the various measures of CEO power. The firm characteristic variables are not significant except for size squared for the first three measures of CEO power. As with the prior results, we see that large bank performance is significantly negatively related to real estate loan exposure. Holding bank characteristics constant, independence and independent financial expertise is negatively related to performance. None of the CEO power variables are related to firm performance. To directly test whether the powerful CEOs are driving the negative relationship between financial expertise among independent directors and performance during the crisis, we 23

interact the proxies of CEO power with the variable measuring financial expertise of the board. If this interaction term is negative and statistically significant, it would lead us to conclude in favor of the rubber stamping hypothesis. However, the interaction term is not significant in all four specifications of CEO power. The interpretation of these finding is that powerful CEOs are not simply getting independent financial experts on the board to rubber stamp high risk taking objectives III.D TARP One potential confounding event that occurred during our sample period is the establishment of the Troubled Asset Relief Program (TARP). Thus, we investigate the effects TARP might have had on the banks in our sample. During the financial crisis, the federal government provided funds to financial institutions and other firms in order to alleviate the risk of systemic failure within the financial system. Though these firms were not in default, these funds were made available by the government to those that needed them. Many banks received TARP money. 18 Although the great majority of the TARP funds were disbursed after the end of our sample period, the exercise is still relevant given the potential anticipation of TARP at the tail end of our sample. We examine the marginal probabilities from logistic regressions estimating the probability of receiving TARP funds. The dependent variable equals one if the bank receives TARP money and zero otherwise. The independent variables include many firm and board characteristics measured at year-end 2006. Not surprisingly, the results (not shown) imply that 18 We obtain data on TARP recipients from http://bailout.propublica.org/main/list/index. This website reports the list of TARP recipients, the amount of TARP money committed, disbursed and returned. Using this website, we identify 160 of the sample firms which received TARP funds. 24

banks with lower equity capital ratios going into the crisis are more likely to receive TARP funding. Additionally, larger boards and more independent boards are associated with increases in the likelihood of receiving TARP funds. Financial expertise is never significant in explaining the probability of receiving TARP funds. These results suggest that TARP-related effects are not driving our main results. IV. Robustness In this section, we provide a series of robustness tests for our main findings. Our robustness results are shown in Table VIII and Table IX. In these tables, we only present the variables associated with independence and financial expertise for brevity since the control variables used are the same as those presented throughout the paper. IV.A All financial institutions Our discussion has focused so far on commercial banks. In this section, we expand our results on independence and financial expertise to the full sample of financial institutions that includes all publicly-traded commercial banks, savings and loan associations (S&Ls) and investment banks. The latter group has been at the center of the crisis and has potentially received the most attention from regulators, the media and researchers. In this extended sample, commercial banks still comprise 75% of all banks with S&Ls make up 23% and investment banks 2%. [Insert Table VIII here] 25

Selected regression results for the sample of all financial institutions are shown in Panel A and B of Table VIII. The results show that financial expertise among independent directors is associated with worse performance during the crisis across all financial institutions (Panel A), which is consistent with our earlier results on commercial banks. As in the main section of our study, the results are stronger for the subset of large financial institutions. For the risk regressions (Panel B), the results show that the pre-crisis market-based measure of risk taking were not significantly higher for large financial institutions that had more financial expertise among their independent directors. Overall, we can conclude that our main performance results extend to the broader sample of financial institutions that include commercial banks, S&Ls and investment banks. IV.B Excluding the five largest U.S. banks Our main sample is comprised of all commercial banks with at least one billion dollars in total assets. As previously noted, there is still considerable heterogeneity in bank size among this sample. We have dealt with this heterogeneity parametrically by including linear and nonlinear size controls in our regressions and non-parametrically using the propensity score matching approach. These approaches may still not capture unobserved differences between smaller commercial banks and the likes of Citigroup, Bank of America or JP Morgan Chase. The largest commercial banks differ in terms of risk taking behavior, market exposure and systemic risk. It is thus important to test the robustness of our results to the exclusion of this set of banks. Specifically, we exclude Citigroup, Bank of America, JP Morgan Chase, Wachovia, and Wells Fargo. 26

We rerun our main tests on banks excluding the largest by total assets at year end 2006 with the results shown in Panel C and D of Table VIII. Results in Panel C show that financial expertise among independent directors is associated with worse performance during the crisis, and the association is stronger for the subset of the large banks. In this case, the set of large banks is comprised of banks above the median total assets at fiscal year-end 2006, excluding the five largest banks. For the risk regressions (Panel D), the results show that the pre-crisis marketbased measure of risk taking was significantly higher for large banks that had more financial expertise among their independent directors, consistent again with our sample of all commercial banks. Overall, we can conclude that our main results extend to the sample of commercial banks that excludes the largest publicly-traded banks. IV.C Other measures of financial expertise and independence In this section, we investigate the impact of using other definitions of independence and financial expertise. Thus far, the definitions of independence are taken from those issued by the exchanges which are in accordance with federal mandates issued through Sarbanes-Oxley. We also follow the definition of financial expertise used in the existing literature. Table IX gives the results for the main set of regressions when examining alternate definitions of both independence and financial expertise. Our alternate measure of board independence in Panel A, called %Truly Indep, takes into consideration outside relationships between the current CEO and independent directors. Using the BoardEx database, we identify independent directors who have a prior or existing relationship with the CEO beyond that of being on the same board. These external ties might include social (e.g. charities), educational or other business relationship. An independent director is considered truly independent if he/she 27

has no relationship with the President/CEO other than being currently on the same bank board (see Fracassi and Tate (2010) for a similar construct). As a consequence, our financial expertise measure is also modified and becomes the percentage of truly independent financial experts computed from the pool of truly independent directors. In terms of financial expertise, we recompute all of our main specifications using a stricter definition by considering independent directors to be experts only if they have held or currently hold executive positions within financial institutions (SIC codes 6000-6999). Results are shown in Panel B of Table IX. [Insert Table IX here] Since all of our main results are driven by the large banks, we concentrate on this subset in this section. In Panel A, the truly independent measure of independence yields qualitatively similar results to our base measure of independence. Large banks with more independent boards did not perform any worse than other banks while those with more financial expertise among independent directors performed worse during the crisis. Banks with more truly independent boards took on less risk leading up to the crisis while banks with more financial expertise among truly independent directors took on more risk using the alternative measure. Finally, in Panel B of Table IX we use the more restrictive definition of financial expertise where we define an independent director as a financial expert only if they held or is currently holding an executive position within a financial institution (SIC code 6000-6999). With this definition, we see that the crisis results hold, in particular with the augmented set of controls. Furthermore, the total risk results are more significant when financial expertise is defined in this manner. The Tier-1 capital ratios results are, however, no longer statistically 28

significant. Overall, the main results of the effect of independence and financial expertise among independent directors is robust across several alternative definitions of these variables. IV.D Other measures of risk There are many activities that a bank can undertake to increase its risk profile in the runup to the crisis. We have shown that higher levels of financial expertise among independent directors were associated with lower Tier-1 capital and with higher levels of real estate exposure. We examine several other measures of bank activities, such as the amount of non-deposit shortterm financing, and the exposure to on- and off-balance sheet asset backed securities related activities. In these regressions (not shown for brevity), the levels of independence and financial expertise are not significantly associated with these variables. V. Conclusion In the aftermath of the 2007-2008 financial crisis, banks and other financial institutions were accused of engaging in excessive risk taking. The composition of the board of directors should be a reliable proxy of how well the board can process information provided by insiders and advise as well as monitor the bank s risk taking practices in the best interests of its shareholders. This paper examines how the presence of financial experts among independent directors relates to risk taking and performance of commercial bank holding companies during the period from 2003 to 2008, which includes the 2007-2008 financial crisis. We document that the level of financial expertise among independent directors is positively related to risk taking at the onset of the financial crisis using both market-based and 29

balance sheet-based risk measures. These results are primarily driven by large banks in our sample. We show that, as a consequence of the heightened risk profile, stock performance was worse during the crisis for large banks with more independent financial expertise. Additionally, there is weak evidence that the higher risk taking of banks with more independent financial expertise is related to better stock performance in the period leading up to the crisis. Our results are robust to alternative measures of independence and financial expertise as well as including S&Ls and investment banks in the sample and excluding the largest U.S. banks from our sample. The Bank of International Settlements (BIS), in their 2006 report Enhancing Corporate Governance for Banking Organisations, stressed that banks should have independent directors and that these directors should have sufficient knowledge of the main financial activities of the bank to enable effective governance and oversight. Overall, our results challenge the regulators view that more financial expertise on the boards of banks would unambiguously lower their risk profile. In particular, the presence of financial experts among independent directors is related to more risk taking in the run-up to the crisis. This positive association with risk-taking was not penalized by the markets prior to the crisis but was penalized by the market when the crisis hit. For the large banks, more financial experts among independent directors led to significant underperformance. We do not find evidence for a reverse causality channel explanation of our results, whereby a powerful CEO would choose a higher risk profile and select independent financial experts to rubber stamp his/her strategy. Our results could be explained by the fact that independent financial experts, with a fiduciary duty to shareholders, understand the residual nature of the equity claims and will generally favor more risk taking. Another plausible explanation could be that external financial 30

experts are more willing to let their bank participate in more risk-taking activities due to their familiarity and understanding of complex financial instruments. Both explanations are important considerations for policy makers as they examine governance reforms of the financial sector. 31

References Acharya, V., T. Cooley, M. Richardson, and I. Walter, 2010, Regulating Wall Street: The Dodd- Frank Act and the New Architecture of Global Finance, New York University Stern School of Business, John Wiley and Sons. Adams, R., 2009, Governance and the financial crisis, unpublished working paper, University of Queensland, Brisbane, Australia. Adams, R., B. Hermalin, and M. Weisbach, 2010, The Role of Boards of Directors in Corporate Governance: A Conceptual Framework & Survey, Journal of Economic Literature 48, 58-107. Adams, R. and H. Mehran, 2003, Is corporate governance different for bank holding companies?, Economic Policy Review 9, 123-142. Adams, R. and H. Mehran, 2008, Corporate Performance, Board Structure, and Their Determinants in the Banking Industry, Federal Reserve Bank of New York Staff Reports, no. 330. Adams, R. and H. Mehran, 2012, Bank board structure and performance: Evidence for large bank holding companies, Journal of Financial Intermediation 21, 243-267. Aebi, V., G. Sabato, M. Schmid, 2011, Risk Management, Corporate Governance, and Bank Performance in the Financial Crisis, Journal of Banking and Finance, forthcoming. Aggarwal, R., I. Erel, R. Stulz, and R. Williamson, 2009, Differences in Governance Practices between U.S. and Foreign Firms: Measurement, Causes, and Consequences, Review of Financial Studies 22, 3171-3209. Bebchuk, L., A. Cohen, and H. Spamann, 2010, The Wages of Failure: Executive Compensation at Bear Stearns and Lehman 2000-2008, Yale Journal on Regulation 27, 257-282. Bebchuk, L. A. and H. Spamann, 2010, Regulating Bankers' Pay, Georgetown Law Journal 98, 247-287. Beltratti, A. and R. M. Stulz, 2010, The credit crisis around the globe: Why did some banks perform better?, unpublished working Paper, Dice Center WP 2010-05, The Ohio State University. Cheng, S., 2008, Board size and the variability of corporate performance, Journal of Financial Economics 87, 157-176. Coles, J., N. Daniel, and L. Naveen, 2008, Boards: Does One Size Fit All?, Journal of Financial Economics 87, 329 356. 32

Cornett, M.M., J.J. McNutt, and H. Tehranian, 2010, The financial crisis, internal corporate governance, and the performance of publicly-traded U.S. bank holding companies, unpublished working paper, Boston College. Dahya, J., McConnell, J., and N. Travlos, 2002, The Cadbury Committee, corporate performance and top management turnover, Journal of Finance 57, 461-483. Drucker, S. and M. Puri, 2005, On the Benefits of Concurrent Lending and Underwriting, Journal of Finance 60, 2763-2799. Erkens, D., M. Hung, and P. Matos, 2009, Corporate Governance in the Recent Financial Crisis: Evidence from Financial Institutions Worldwide, unpublished working paper, Marshall School of Business, USC. Ellul, A. and V. Yerramilli, 2010, Stronger Risk Controls, Lower Risk: Evidence from U.S. Bank Holding Companies, unpublished working paper, Kelley School of Business, Indiana University. Fahlenbrach, R., B. Minton, and C. Pan, 2011, Former CEO Directors: Lingering CEOs or Valuable Resources?, Review of Financial Studies 24, 3486-3518. Fahlenbrach, R. and R. M. Stulz, 2011, Bank CEO Incentives and the credit crisis, Journal of Financial Economics 99, 11-26. Ferreira, D., T. Kirchmaier, and D. Metzger, 2010, Boards of Banks, Finance Working Paper N. 289/2010, ECGI. Fernandes, N. and E. Fich, 2009, Does Financial Experience Help Banks during Credit Crises?, unpublished working paper, IMD-Lausanne, Switzerland. Fich, E. and A. Shivdasani, 2006, Are Busy Boards Effective Monitors?, Journal of Finance 61, 689-724. Fracassi, C. and G. Tate, 2010, External Networking and Internal Firm Governance, unpublished working paper, UCLA. Guerrera, F. and P. Thal-Larsen, 2008, Gone by the Board: Why the directors of big banks failed to spot credit risks, Financial Times, June 26 th. Güner, A.B., Malmendier. U., and G. Tate, 2008, Financial expertise of directors, Journal of Financial Economics 88, 323-354. Harris, M. and A. Raviv, 2008, A Theory of Board Control and Size. Review of Financial Studies 21, 1797-832. 33

Hashagen, J., N. Harman, and M. Conover, 2009, Never again? Risk management in banking beyond the credit crisis. KPMG International. Hermalin, B., and M. Weisbach, 1998, Endogenously chosen boards of directors and their monitoring of the CEO, American Economic Review 88, 96-118. Keeley, M., 1990, Deposit Insurance, Risk, and Market Power in Banking, American Economic Review 80, 1183-200. Kirkpatrick, G., 2009, The corporate governance lessons from the financial crisis, report, OECD, Paris, France. Landier, Augustin, David Sraer, and David Thesmar, 2007, Bottom-Up Corporate Governance, unpublished working paper, HEC Paris. Linck, J., J. Netter, and T. Yang, 2008, The Determinants of Board Structure, Journal of Financial Economics 87, 308-328. Marcus, A. 1984, Deregulation and Bank Financial Policy, Journal of Banking and Finance 8, 557-565. Merton, R. C., 1977, An Analytic Derivation of the Cost of Deposit Insurance and Loan Guarantees, Journal of Banking and Finance 1, 3-11. Pathan, S., 2009, Strong boards, CEO power and bank risk-taking, Journal of Banking and Finance 33, 1340-1350. Strebel, R., 2009, Time to bring real shareholders back on board, Financial Times, February 12 th. Walker, D., 2009, A Review of Corporate Governance in UK banks and other Financial Industry Entities: Final Recommendations. http://www.hmtreasury.gov.uk/walker_review_information.htm. 34

Table I: Summary statistics This table shows summary statistics for board governance and other bank characteristics for our sample of U.S. publicly-traded bank holding companies with total assets greater than $1 billion from 2003 to 2008. Panel A presents yearly statistics on key board governance dimensions. Board size is the number of directors on the board. %independent is the percent of directors that are not employed or affiliated with the firm. These two variables are computed using the Boardex database. % fin expert among independent directors is the percent of independent directors who are financial experts. Similar to Güner et al. (2008), an independent director is classified as a financial expert if he or she (i) has held an executive position at a banking institution (Former bank executive), (ii) holds an executive position at a non-bank financial institution (Exec. Of nonbank financials), (iii) holds a finance-related position (e.g. chief financial officer (CFO), accountant, treasurer, VP finance) of a non-financial firm (Finance exec. of non-financials), (iv) holds an academic position in a related field (e.g. Professor of finance, economics or accounting), or (v) works as a hedge fund or private equity fund manager, or venture capitalist (Professional investor). The data on financial expertise of each director were provided by a combination of Boardex data and hand-collected data from annual proxy statements. Panel B presents the correlation matrix between our board characteristics and several other bank characteristics. Log(assets) is the natural logarithm of total book assets. Equity capital ratio is defined as total book equity divided by total assets. We also computed dummy variables for institutions that disclose having a risk management committee (Has Risk Comm.) or chief risk officer (CRO) in annual 10-K filings or proxy statements during the sample period. These data were compiled by hand-collecting all the information on CRO and other similarly-defined risk management executive positions and (non-alm) risk committees within a financial institution, as disclosed by the firm s 10-K and DEF 14-A (proxy statement) each year. Panel A: Yearly summary statistics (mean) on board characteristics Year Number of obs. %Fin. expert. among independent directors Board size %Independent 2003 182 20% 13.3 73% 2004 200 22% 13.1 76% 2005 206 23% 12.6 77% 2006 203 24% 12.6 77% 2007 196 25% 12.5 77% 2008 119 26% 12.4 78% 35

Panel B: Correlation matrix among board and risk governance and bank characteristics Log(Asset) Equity capital ratio CRO Risk committee Board size %Independent directors Log(Asset) 1.00 Equity capital ratio 0.11 1.00 CRO 0.36 0.05 1.00 Risk committee 0.42 0.01 0.35 1.00 Board size 0.35 0.21 0.06 0.09 1.00 %Independent directors 0.14-0.04 0.16 0.16-0.04 1.00 %Fin expert among indep. directors -0.13-0.07 0.05 0.02-0.13 0.01 36

Table II: Total risk and financial expertise This table presents results for OLS panel regressions of market-based measures of risk-taking activities on firm and board characteristics. Total risk is the average annual standard deviation of daily stock returns over the 2004-2006 period. All independent variables are defined in Table I and are measured at year-end 2003. A bank is defined as large for a given year if the book value of its total assets is greater than the median total asset number for the sample of commercial bank during that year. The standard deviation of the error term is estimated using clustering at the firm level. Standard errors are given in brackets. ***, ** and * denote respectively significance at the 1%, 5% and 10% levels. All banks Large Banks (1) (2) (1) (2) Log(assets) -0.713*** -0.685*** -0.843*** -0.853*** [0.183] [0.179] [0.288] [0.279] Log(assets) 2 0.028*** 0.027*** 0.034** 0.036*** [0.009] [0.009] [0.014] [0.013] Equity capital ratio 0.017 0.521 0.384 1.109 [1.642] [1.679] [1.814] [1.802] Board Size -0.016*** -0.014*** -0.014** -0.013** [0.005] [0.005] [0.005] [0.005] Has CRO 0.084 0.074 0.087 0.065 [0.067] [0.064] [0.079] [0.074] Has Risk Comm. 0.002-0.014 0.008-0.010 [0.045] [0.047] [0.054] [0.055] %Independent -0.001-0.034-0.201-0.295 [0.222] [0.226] [0.227] [0.228] %Fin expert among independent directors 0.237* 0.378** [0.136] [0.148] Observations 96 96 76 76 Adjusted R-squared 0.541 0.550 0.462 0.494 37

Table III: Balance sheet risk and financial expertise in 2006 This table presents results for cross-sectional OLS regressions in 2006 for Tier 1 capital ratios (Panel A) and real estate exposure (Panel B) on firm and board characteristics. Real estate loans are computed as all loans secured by real estate relative to total assets. Data on real estate loans and Tier-1 capital ratios are from the FED FRY-9C Call Report forms filed by Bank Holding Companies (BHCs). Net interest margin is a measure of bank profitability used as a control for the Tier 1 capital ratio regressions. All other board and risk monitoring variables are defined in Table I. All independent variables are measured at year-end 2005. A bank is defined as large if the book value of total assets is greater than the median total asset number for the sample of commercial banks at year-end 2005. Robust standard errors corrected for heteroskedasticity are reported in brackets. ***, ** and * denote respectively significance at the 1%, 5% and 10% levels. Panel A: Tier-1 risk-weighted capital ratio All banks Large banks (1) (2) (1) (2) Log(assets) 0.009 0.009-0.010-0.012 [0.009] [0.010] [0.014] [0.014] Log(assets) 2-0.001-0.001 0.000 0.000 [0.000] [0.000] [0.001] [0.001] Net interest margin -0.005* -0.005* -0.006* -0.007* [0.003] [0.003] [0.003] [0.003] Board Size 0.000 0.000-0.000-0.000 [0.000] [0.000] [0.001] [0.001] Has CRO 0.002 0.003 0.004 0.005 [0.004] [0.004] [0.004] [0.005] Has Risk Comm. -0.006-0.005-0.007-0.007 [0.004] [0.004] [0.005] [0.005] %Independent -0.022-0.023-0.001 0.000 [0.018] [0.018] [0.017] [0.016] %Fin expert among independent -0.013-0.019** [0.009] [0.008] Observations 193 193 106 106 Adjusted R-squared 0.133 0.141 0.190 0.227 38

Panel B: Real estate loans All banks Large banks (1) (2) (1) (2) Log(assets) -0.083-0.084-0.191-0.198 [0.074] [0.075] [0.178] [0.181] Log(assets) 2 0.002 0.002 0.007 0.008 [0.004] [0.004] [0.008] [0.008] Has a long-term credit -0.035-0.008 rating -0.033-0.001 [0.043] [0.043] [0.054] [0.055] Board Size 0.002 0.002 0.003 0.003 [0.003] [0.003] [0.004] [0.004] Has CRO 0.013 0.011 0.003-0.002 [0.029] [0.029] [0.038] [0.039] Has Risk Comm. 0.022 0.021 0.033 0.033 [0.030] [0.031] [0.037] [0.037] %Independent -0.111-0.109-0.235-0.245 [0.099] [0.099] [0.152] [0.150] %Fin expert among independent 0.039 0.079 [0.062] [0.089] Observations 193 193 106 106 Adjusted R-squared 0.166 0.163 0.140 0.140 39

Table IV. Propensity Score Matching Analyses This table presents results for propensity-score matching analyses for total risk (Panel A), Tier 1 riskweighted capital ratios (Panel B), and real estate loans (Panel C). A bank is considered part of the treatment group if it has at least one independent financial expert on its board. Total risk is the average annual standard deviation of daily stock returns over the 2004-2006 period. Real estate loans are computed as all loans secured by real estate relative to total assets. Data on real estate loans and Tier-1 capital ratios are from the FED FRY-9C Call Report forms filed by Bank Holding Companies (BHCs). A bank is defined as large if the book value of total assets is greater than the median total asset number for the sample of commercial banks at year-end 2005. ***, ** and * denote respectively significance at the 1%, 5% and 10% levels. Panel A: Total Risk All banks Large banks Bank with at least one independent financial expert 1.365 1.302 Control group 1.217 1.203 Difference 0.147** 0.099 T-stat 2.312 1.502 Panel B: Tier 1 risk-weighted capital ratio Bank with at least one independent financial expert All banks Large banks 0.109 0.107 Control group 0.114 0.115 Difference -0.005-0.008* T-stat -1.167-1.647 Panel C: Real estate loans Bank with at least one independent financial expert All banks Large banks 0.503 0.459 Control group 0.447 0.400 Difference 0.055*** 0.059* T-stat 2.106 1.78 40

Table V: Stock performance during the crisis This table presents results for OLS regressions of firm stock performance on firm and board characteristics and risk monitoring functions. Panel A reports the results for all banks. Panel B reports the results for large banks. A bank is defined as large if its book value of total assets is greater than that of the median firm in the sample for 2006. Stock performance is measured as the nominal cumulative stock return from January 2007 to December 2008. Log(assets) is the natural logarithm of total book assets. Equity capital ratio is defined as total book equity divided by total assets. Beta is computed as the market beta estimated from a market model in which the daily stock returns are explained by the valueweighted market return and the return on the three-month Treasury bill. Stock ret 2006 is the nominal cumulative stock return over 2006. All board and risk monitoring variables are defined in Table I. Data on loans, deposits, real estate loans, and mortgage-backed securities are from the FED FRY-9C Call Report forms filed by Bank Holding Companies (BHCs). Real estate loans are computed as all loans secured by real estate relative to total assets. MBS are mortgage-backed securities held-to-maturity, available-for-sale and held for trading relative to total assets. Short-term financing is non-deposit shortterm financing to total asset. Off-Balance sheet activities is the off balance sheet securitized assets (home equity, credit card, auto loans, etc.) to total assets. All variables are taken as of the fiscal year end 2006. Robust standard errors corrected for heteroskedasticity are reported in brackets. ***, ** and * denote respectively significance at the 1%, 5% and 10% levels. 41

Panel A: All banks (1) (2) (3) (4) Log(assets) 0.256 0.271 0.169 0.191 [0.181] [0.182] [0.182] [0.187] Log(assets) 2-0.014-0.015-0.012-0.013 [0.009] [0.009] [0.010] [0.010] Equity capital ratio -1.433-1.574-0.052-0.213 [1.092] [1.078] [1.044] [1.056] Beta 0.153*** 0.153*** 0.142*** 0.142*** [0.040] [0.039] [0.035] [0.035] Stock ret 2006-0.074-0.031-0.033-0.006 [0.172] [0.170] [0.143] [0.142] Board Size 0.011 0.010 0.007 0.007 [0.008] [0.008] [0.007] [0.007] Has CRO 0.089 0.100* 0.106* 0.112** [0.058] [0.058] [0.055] [0.056] Has Risk Comm. -0.059-0.055-0.042-0.041 [0.083] [0.081] [0.078] [0.077] Total loans -0.088-0.027 [0.415] [0.422] Deposits 0.436 0.369 [0.537] [0.537] Real estate loans -0.849*** -0.885*** [0.256] [0.263] MBS 0.479 0.431 [0.403] [0.408] Short-term financing 0.800 0.784 [0.840] [0.847] Off-balance sheet 0.480 0.509 [0.493] [0.487] % Independent directors -0.487** -0.496** -0.510** -0.527** [0.237] [0.234] [0.228] [0.226] %Fin expert among -0.228* -0.155 independent directors [0.134] [0.132] Observations 198 198 198 198 Adjusted R-squared 0.066 0.076 0.229 0.232 42

Panel B: Large banks (1) (2) (3) (4) Log(assets) 0.592 0.629 0.400 0.506 [0.413] [0.424] [0.372] [0.381] Log(assets) 2-0.029-0.031-0.023-0.029 [0.019] [0.020] [0.018] [0.018] Equity capital ratio -0.871-1.521 0.739 0.019 [1.514] [1.516] [1.510] [1.475] Beta 0.235* 0.227* 0.181 0.182 [0.131] [0.135] [0.112] [0.113] Stock ret 2006 0.033 0.086 0.106 0.132 [0.311] [0.317] [0.290] [0.286] Board Size 0.006 0.009 0.005 0.007 [0.013] [0.013] [0.012] [0.011] Has CRO 0.090 0.120* 0.071 0.092 [0.075] [0.071] [0.072] [0.068] Has Risk Comm. -0.122-0.131-0.078-0.094 [0.084] [0.080] [0.076] [0.074] Total loans 0.054 0.173 [0.625] [0.621] Deposits -0.129-0.280 [0.724] [0.658] Real estate loans -1.244** -1.294*** [0.498] [0.469] MBS 0.443 0.329 [0.528] [0.571] Short-term financing -0.559-0.784 [1.049] [1.023] Off-balance sheet 0.423 0.565 [0.518] [0.504] % Independent directors -0.625* -0.521-0.753** -0.731** [0.331] [0.321] [0.326] [0.316] %Fin expert among -0.593*** -0.503*** independent directors [0.186] [0.175] Observations 101 101 101 101 Adjusted R-squared 0.033 0.116 0.217 0.274 43

Table VI: Stock performance over the pre-crisis period This table presents results for OLS regressions of firm stock performance over the pre-crisis period on firm and board characteristics and corresponding propensity-score matching analysis. Panel A and B report the results for all banks. Panel C and D report the results for large banks. A bank is defined as large if its book value of total assets is greater than the median total asset number for the sample of commercial banks for 2003. Stock performance is measured as the nominal cumulative stock return from January 2003 to December 2006. Log(assets) is the natural logarithm of total book assets. Equity capital ratio is defined as total book equity divided by total assets. Beta is computed as the market beta estimated from a market model in which the daily stock returns are explained by the value-weighted market return and the return on the three-month Treasury bill. All board and risk monitoring variables are defined in Table I. All variables are taken as of the fiscal year-end 2003. Robust standard errors corrected for heteroskedasticity are reported in brackets. ***, ** and * denote respectively significance at the 1%, 5% and 10% levels. 44

Panel A: All banks (1) (2) (3) (4) Log(Assets) -1.073* -0.973* -0.775-0.666 [0.553] [0.548] [0.518] [0.515] Log(Assets) squared 0.052* 0.049* 0.039 0.035 [0.026] [0.026] [0.025] [0.025] Book equity 0.315 1.504-2.242-0.965 [2.038] [2.041] [2.984] [2.919] Beta 0.066-0.054 0.028-0.106 [0.221] [0.222] [0.269] [0.265] Board Size -0.008-0.006-0.004-0.002 [0.012] [0.012] [0.012] [0.012] Has CRO -0.092-0.105-0.050-0.056 [0.098] [0.097] [0.120] [0.112] Has Risk Comm. -0.111-0.157* -0.170-0.215* [0.081] [0.093] [0.105] [0.117] Total loans -0.452-0.494 [0.624] [0.646] Total deposits 0.198 0.383 [0.894] [0.758] Real estate loans 0.135 0.061 [0.551] [0.559] Mortgage-backed securities -0.965* -1.071* [0.558] [0.557] Short-term financing -1.982-1.860* [1.236] [1.091] Off-balance sheet commitments -0.442-0.535 [0.444] [0.466] %independent directors 0.058-0.050 0.142 0.003 [0.272] [0.282] [0.292] [0.313] %Fin expert among 0.644** 0.714** independent directors [0.280] [0.307] Observations 100 100 100 100 Adjusted R-squared 0.034 0.066 0.042 0.085 45

Panel B: Large banks (1) (2) (3) (4) Log(Assets) -0.727* -0.730* -0.486-0.483 [0.404] [0.415] [0.405] [0.410] Log(Assets) squared 0.036* 0.037* 0.026 0.026 [0.019] [0.020] [0.019] [0.019] Book equity 1.789 2.210 1.425 1.879 [2.327] [2.383] [2.195] [2.313] Beta 0.185 0.142 0.218 0.146 [0.184] [0.191] [0.185] [0.192] Board Size -0.004-0.004-0.003-0.003 [0.009] [0.009] [0.009] [0.009] Has CRO -0.063-0.072-0.017-0.032 [0.095] [0.096] [0.106] [0.106] Has Risk Comm. -0.063-0.077-0.063-0.083 [0.079] [0.082] [0.093] [0.098] Total loans -0.866* -0.896* [0.474] [0.476] Total deposits 0.853 0.796 [0.573] [0.551] Real estate loans 0.418 0.368 [0.458] [0.459] Mortgage-backed securities -0.666-0.691* [0.412] [0.413] Short-term financing -0.091-0.225 [0.896] [0.913] Off-balance sheet commitments -0.395-0.429 [0.425] [0.424] %independent directors -0.158-0.221-0.143-0.216 [0.179] [0.183] [0.192] [0.196] %Fin expert among 0.256 0.317 independent directors [0.220] [0.245] Observations 80 80 80 80 Adjusted R-squared -0.008-0.008 0.013 0.02 46

Table VII: Stock performance during the crisis and CEO power This table presents OLS regression coefficients of firm stock performance during the crisis on firm and board characteristics for a sample of large banks. A bank is defined as large if its book value of total assets is greater than the median total asset number for the sample of banks in 2006. Stock performance is measured as the nominal cumulative stock return from January 2007 to December 2008. Log(assets) is the natural logarithm of total book assets. Equity capital ratio is defined as total book equity divided by total assets. Beta is computed as the market beta estimated from a market model in which the stock returns are explained by the value-weighted market return and the return on the three-month Treasury bill. Stock ret 2006 is the nominal cumulative stock return over 2006. All board and risk monitoring variables are defined in Table I. Data on loans, deposits, real estate loans, and mortgage backed securities are from the FED FRY-9C Call Report forms filed by Bank Holding Companies (BHCs). Real estate loans are computed as all loans secured by real estate relative to total assets. MBS are mortgage-backed securities held-to-maturity, available-for-sale, and held for trading relative to total assets Short-term financing is non-deposit short-term financing to total asset. Off-Balance sheet activities is the off balance sheet securitized assets (home equity, credit card, auto loans, etc.) to total assets. In regression (1) the proxy for CEO power is an indicator variable equal to one if the CEO is also the Chairman of the Board and zero otherwise. In regression (2) the proxy for CEO power is the percentage of board members whose tenure on the board exceeds that of the current CEO so the indicator variable is equal to one if the bank falls below the median on this measure of relative CEO tenure. In regression (3) the proxy for CEO power is the percentage of independent directors with greater tenure and the indicator variable we derive from this measure takes the value of one for all firms below the median of the commercial bank sample. In regression (4) the proxy for CEO power is the percentage of independent financial experts whose tenure exceeds that of the CEO and we also derive a CEO strength variable from it that takes the value of one for all banks that fall below the median value on that dimension. All variables are taken as of the fiscal year end 2006. Robust standard errors corrected for heteroskedasticity are reported in brackets. ***, ** and * denote respectively significance at the 1%, 5% and 10% levels.

(1) (2) (3) (4) Log(assets) 0.541 0.563 0.611 0.530 [0.390] [0.382] [0.382] [0.386] Log(assets) 2-0.032* -0.032* -0.034* -0.031* [0.019] [0.018] [0.018] [0.018] Equity capital ratio -0.114 0.006-0.031-0.046 [1.491] [1.441] [1.430] [1.445] Beta 0.159 0.177 0.178 0.162 [0.120] [0.117] [0.118] [0.124] Stock ret 2006 0.148 0.191 0.199 0.212 [0.273] [0.293] [0.291] [0.303] Board Size 0.006 0.007 0.006 0.006 [0.011] [0.011] [0.011] [0.011] Has CRO 0.084 0.100 0.093 0.106 [0.073] [0.069] [0.070] [0.072] Has Risk Comm. -0.095-0.128-0.134* -0.116 [0.077] [0.080] [0.078] [0.076] Total loans 0.042 0.061 0.053 0.029 [0.592] [0.620] [0.615] [0.656] Deposits -0.596-0.450-0.483-0.418 [0.693] [0.679] [0.676] [0.685] Real estate loans -1.333*** -1.233** -1.215** -1.242** [0.481] [0.476] [0.477] [0.473] MBS 0.399 0.516 0.516 0.480 [0.541] [0.612] [0.592] [0.612] Short-term financing -1.603-1.268-1.356-1.269 [1.038] [1.072] [1.072] [1.164] Off-balance sheet 0.648 0.604 0.602 0.627 [0.510] [0.489] [0.476] [0.521] % Independent directors -0.799** -0.798** -0.751** -0.809** [0.319] [0.337] [0.344] [0.330] %Fin expert among -0.202-0.672** -0.790** -0.574* independent directors [0.336] [0.309] [0.327] [0.335] CEO power proxy 0.190-0.051-0.089-0.017 [0.123] [0.111] [0.105] [0.129] CEO power proxy * %Fin exp -0.480 0.254 0.423 0.106 [0.389] [0.362] [0.363] [0.446] Observations 98 98 98 98 Adjusted R-squared 0.286 0.267 0.274 0.263 48

Table VIII: Sample-related robustness tests Each panel of this table replicates the performance in crisis regressions (Table IV) and the market-based measures of risk taking in the run-up to the crisis (Table II) on the extended sample of all publicly-traded commercial banks, savings and loan associations (S&Ls) and investment banks in Panel A and B and the sample of publicly-traded commercial banks, excluding the five largest commercial banks (Citigroup, Bank of America, JP Morgan Chase, Wachovia, and Wells Fargo) in Panel C and D. A bank-type indicator variable is added to each specification of Panel A and B. Robust standard errors corrected for heteroskedasticity are reported in brackets. ***, ** and * denote respectively significance at the 1%, 5% and 10% levels. Panel A: Crisis results on broader set of financial institutions Performance All Financial Institutions Large Financial Institutions Controls Yes Yes % fin experts among independent directors -0.227* -0.443*** [0.119] [0.165] Observations 272 137 Adjusted R-squared 0.071 0.117 Panel B: Pre-crisis results on broader set of financial institutions - Total risk All financial Institutions Large Financial Institutions Controls Yes Yes % fin experts among independent directors 0.198 0.260 [0.138] [0.160] Observations 131 105 Adjusted R-squared 0.507 0.433 Panel C: Crisis results on commercial banks excluding largest banks Performance All Banks Large Banks Controls Yes Yes % fin experts among independent directors -0.242* -0.610 *** [0.135] [0.187] Observations 194 99 Adjusted R-squared 0.083 0.141 Panel D: Pre-crisis results on commercial banks excluding largest banks - Total risk All Banks Large Banks Controls Yes Yes % fin experts among independent directors 0.202 0.327** [0.143] [0.147] Observations 92 74 Adjusted R-squared 0.520 0.510

Table IX: Robustness of board characteristics definitions Each panel of this table replicates the performance in crisis regressions (Table IV) and the market-based and balance sheet-based measures of risk taking in the run-up to the crisis (Table II and III) for the sample of large banks using alternate definitions for board independence and financial expertise. In Panel A, we use a more restrictive definition of independence. Among the declared independent board members, we compute the percentage of independent directors on the board who do not have any present or past social or economic ties with the current CEOs beyond being on the same bank board. This variable is computed using Boardex database and matching up a given director with the current CEO against all educationrelated, charity-related and business-related institutions. In Panel B, we use a more restrictive definition of financial expertise relative to Güner et al. (2008). Specifically, an independent director is defined as a financial expert only if he has been or currently is an executive within the financial industry (SIC code between 6000 and 6999). Robust standard errors corrected for heteroskedasticity are reported in brackets. ***, ** and * denote respectively significance at the 1%, 5% and 10% levels. Panel A: More restrictive definition of independence Crisis regressions: Large banks Base controls Augmented set of controls (1) (2) (1) (2) %truly independent directors -0.081-0.030-0.068-0.032 [0.139] [0.135] [0.132] [0.132] % fin experts among truly independent directors -0.412*** -0.352** [0.147] [0.141] Observations 101 101 101 101 Adjusted R-squared 0.004 0.067 0.175 0.220 Total risk in pre-crisis: Large banks (1) (2) Controls Yes Yes % truly independent directors -0.171-0.241** [0.106] [0.108] % fin experts among truly independent directors 0.268** [0.112] Observations 76 76 Adjusted R-squared 0.475 0.493 Balance sheet Tier 1 risk-weighted capital ratio: Large banks (1) (2) Controls Yes Yes %truly independent directors -0.014** -0.013* [0.007] [0.007] % fin experts among truly independent directors -0.007 [0.007] Observations 106 106 Adjusted R-squared 0.227 0.225 50

Panel B: More restrictive definition of financial expertise Crisis regressions: Large banks Base controls Augmented set of controls (1) (2) (1) (2) %independent directors -0.625* -0.551-0.753** -0.739** [0.331] [0.342] [0.326] [0.331] %fin experts among independent directors -0.459** -0.417** [0.233] [0.185] Observations 101 101 101 101 Adjusted R-squared 0.033 0.061 0.217 0.239 Total risk in pre-crisis: Large banks (1) (2) Controls Yes Yes %independent directors -0.201-0.318 [0.227] [0.227] %fin experts among independent directors 0.575** [0.221] Observations 76 76 Adjusted R-squared 0.462 0.510 Balance sheet Tier 1 risk-weighted capital ratio: Large banks (1) (2) Controls Yes Yes % independent directors -0.001-0.000 [0.017] [0.017] %fin experts among independent directors -0.014 [0.010] Observations 106 106 Adjusted R-squared 0.190 0.195 51

Figure 1: Composition of financial expertise This figure provides the average composition of financial expertise among independent directors of the board for publicly traded bank holding companies with total assets greater than $1Bn over the period 2003-2008. Similar to Güner et al. (2008), an independent director is classified as a financial expert if he or she (i) has held an executive position at a banking institution (Former bank executive), (ii) holds an executive position at a non-bank financial institution (Exec. Of non-bank financials), (iii) holds a financerelated position (e.g. chief financial officer (CFO), accountant, treasurer, VP finance) of a non-financial firm (Finance exec. of non-financials), (iv) holds an academic position in a related field (e.g. Professor of finance, economics or accounting), or (v) works as a hedge fund or private equity fund manager, or venture capitalist (Professional investor). 52