Governance and the Financial Crisis

Authors


  • I thank Denis Sosyura, participants at the 2009 Copenhagen Corporate Governance conference, the 2009 Helsinki PhD conference, and the 2010 Asian FA meetings for helpful comments.

Renée Adams

University of New South Wales and ECGI

Sydney

NSW 2052

Australia

renee.adams@unsw.edu.au

Abstract

Should boards of financial firms be blamed for the financial crisis? Using a large sample of data on nonfinancial and financial firms for the period 1996–2007, I document that the governance of financial firms is, on average, not obviously worse than in nonfinancial firms. In fact, using simple governance scores and governance indices as measures, banks and nonbank financial firms generally appear to be better governed than nonfinancial firms. I also document that bank directors earned significantly less compensation than their counterparts in nonfinancial firms and banks receiving bailout money had boards that were more independent than in other banks. My results suggest that measures of governance that have been the focus of recent governance policies are insufficient to describe governance failures attributed to financial firms. Moreover, recent governance reforms may have to shoulder some of the blame placed on boards of financial firms.

I. Introduction

Based on measures of world industrial output, world trade and stock markets, Eichengreen and O'Rourke (2009) argue that the current financial crisis may be worse than the Great Depression on a global scale. Perhaps no one would have been surprised if a crisis of this magnitude originated in an emerging market. Bordo and Eichengreen (2003) provide evidence that most financial crises occur in emerging markets. They describe that there were 139 financial crises between 1973 and 1997, 95 of which occurred in emerging market countries.

There are many reasons why investors may lose confidence in emerging markets. If such markets are characterized by weak institutions and poor firm-level governance, then capital outflows and stock market crashes may occur. This is what Johnson et al. (2000), among others, argue this happened in the Asian crisis of 1997–1998. But the current financial crisis originated in the United States, a country that is commonly held up as a role model in terms of institutional strength and good governance. For example, the United States achieves the highest score on La Porta et al. (1998)'s anti-director rights index, which measures how well the legal system protects minority shareholders against managers or dominant shareholders.1 In addition, La Porta et al. (1997) show that common law countries, such as the United States, and countries that score higher on their measure of anti-director rights have more developed capital markets. The explanation is that in countries with better protection of shareholders, financiers are willing to invest more money and on better terms for entrepreneurs.

Not only is the United States seen as having relatively strong legal institutions, but recent regulation designed to strengthen firm-level governance, the Sarbanes–Oxley Act of 2002 (SOX) and new exchange listing requirements at the NYSE and Nasdaq, have served as models for governance reform around the world. I document in Section 2 that 16 countries published new or revised important governance codes or passed laws on governance in 2003 alone. Moreover, significantly more countries instituted governance reforms after 2002 than in the preceding 10 years.

The Sarbanes–Oxley Act and the new listing requirements were a reaction to a series of dramatic corporate and accounting scandals including those at Enron, Tyco, and Worldcom. Much of the blame for these scandals was put on boards of directors. For example, in its report on Enron's collapse, the US Senate argued that by not questioning management about the complicated financial transactions Enron was engaging in, the board had failed in its fiduciary duties to shareholders (US House 2002). Accordingly, both SOX and the NYSE and Nasdaq listing requirements contain a series of provisions designed to strengthen board oversight of management in publicly traded firms.

Currently, for example, a company listed on the NYSE would have to have a majority of independent directors (new NYSE listing standard), an independent audit committee consisting of at least three members (NYSE) and a financial expert or a reason not to have a financial expert (SOX), a completely independent nominating/corporate governance committee (new NYSE listing standard), a completely independent compensation committee (new NYSE listing standard), regularly scheduled meetings of the nonmanagement directors (new NYSE listing standard), and a yearly meeting of the independent directors (new NYSE listing standard). Nasdaq's listed companies are subject to similar requirements, although they do not have to have a separate compensation or nominating committee. In addition, SOX, NYSE, and Nasdaq have tightened the definition of independent director.2

To a certain extent, the fact that the United States is considered to have strong investor protection and good governance may help explain why the financial crisis was predicted by so few. But it raises the question whether and to what extent governance can be considered to be a cause of the financial crisis. The resignations of high-profile finance executives, e.g., Stan O'Neal at Merrill Lynch, Charles Prince at Citigroup, and Marcel Ospel at UBS, and the recommendations by proxy advisors against the reelection of the board at Citigroup, among others (see, e.g., Moyer 2008), are direct evidence that boards are, at least partly, being blamed for the crisis.

The OECD Steering Group on Corporate Governance goes further. It argues that weak governance is a major cause of the financial crisis (Kirkpatrick 2009). It places much of the blame on board failures in financial firms, in particular,3 and has launched an action plan to improve corporate governance.4 Although the UK generally also scores highly on measures of investor protection, bank governance in the UK is also, at least partly, being blamed for the financial crisis. As a result, Sir David Walker was commissioned to recommend measures to improve board-level governance at banks to the government (Walker 2009). A measure of the importance of this review is that it serves as a basis for the 2010 UK Governance code.5

How can it be that governance problems still exist in the United States despite strong shareholder protection mechanisms and recent governance reforms? Can boards of financial firms be to blame for the crisis when publicly traded financial firms have to abide by the same governance requirements in SOX and the listing rules as nonfinancial firms? This article tries to shed more light on the extent to which the crisis can be attributable to bad financial firm governance, in particular board structure and incentives. Both because the financial crisis originated in the United States and because of data limitations, this article focuses on publicly traded financial firms in the United States. Because financial firms around the world have different activities and structures and face different regulatory constraints, providing an overview of financial firm governance across multiple countries is a complex task that is beyond the scope of this article.6

In Section 2, I briefly describe the governance environment prior to the crisis. In Section 3, I discuss what one might expect a well-governed financial firm to look like. In Section 4, I examine whether recent regulatory reform in the United States had an impact on firms' governance structures and compare measures of governance structure across financial and nonfinancial firms in the United States. In Section 5, I discuss whether bad governance contributed to the crisis. I discuss some lessons for the future in Section 6.

II. The Governance Environment Prior to the Crisis

Although the crisis originated in the United States, boards are being blamed for the crisis around the world. Thus, it is important to understand the state of corporate governance around the world prior to the crisis. To do this, I collected data on important governance policies around the world. I define important policies to be governance codes that operate on a comply-or-explain basis or that are clearly national codes even if they are voluntary, as well as governance laws and regulations. The policies consist of both new policies and important updates. I obtain data on codes from the ECGI code database at http://www.ecgi.org. I supplemented these data with data on codes and laws and regulations concerning governance from eStandardsForum (2011). I collected additional data on laws related to governance from Cornelli et al. (2010). For countries not in the ECGI database or not covered by eStandardsForum, I conducted a Google search to identify governance policies. I end with a sample of 171 governance policies that were introduced or had a major update between 1993 and 2010 for 195 countries officially recognized by the US Department of State as of 2011, plus Hong Kong and Taiwan. The average number of policies per year during this time period was 9.33. Figure 1 shows the plot of the number of governance policies per year. What is noticeable from the plot is that the number of governance codes is higher in every year after the introduction of SOX and the listing standards in the United States than before. The differences pre- and post-US reforms are also statistically significant at greater than the 1% level.

Figure 1.

Number of important country-level governance policies per year 1993–2010.

Figure 1 shows the number of important country-level governance policies per year between 1993 and 2010. I define important policies to be codes that operate on a comply-or-explain basis or that are clearly national codes even if they are voluntary, as well as governance laws and regulations. The policies consist of both new policies and important updates. The data are for 195 countries officially recognized by the US Department of State as of 2011 plus Hong Kong and Taiwan. Data on policies are from http://www.eStandardsForum.org, the ECGI code database, Cornelli et al. (2010), and a Google search (see Adams 2011). The vertical lines for years 2002 and 2008 frame the period between the governance reform movement in the United States, embodied in SOX and revised exchange listing rules, and the financial crisis.

Figure 2 shows the plot of the percentage of countries strengthening board-level governance standards by introducing or revising a code per year. The percentage started at 0.508% in 1993 and ended with 43.147% in 2010. In 2007, the percentage of countries strengthening governance standards was 39.086%. The differences pre- and post-US reforms are statistically significant at greater than the 1% level.

Figure 2.

Percentage of countries increasing governance standards 1993–2010.

Figure 2 shows the fraction of countries introducing or revising important country-level governance policies between 1993 and 2010. I define important policies to be codes that operate on a comply-or-explain basis or that are clearly national codes even if they are voluntary, as well as governance laws and regulations. The policies consist of both new policies and important updates. The data are for 195 countries officially recognized by the US Department of State as of 2011 plus Hong Kong and Taiwan. Data on policies are from http://www.eStandardsForum.org, the ECGI code database, Cornelli et al. (2010 and a Google search (see Adams 2011). The vertical lines for years 2002 and 2008 frame the period between the governance reform movement in the US, embodied in SOX and revised exchange listing rules, and the financial crisis.

As I describe in the Introduction, governance standards in the United States were quite high just prior to the financial crisis. As Figures 1 and 2 indicate, the governance environment was strengthening globally as well. The question how governance could have caused the crisis in the United States is therefore a more general question: How can governance be blamed worldwide given that governance standards were strengthening globally?

One possibility is that financial firms did not abide by the governance standards. I examine this issue in more detail in Section 4. Another possibility is that financial firms complied with some aspects of governance standards but were so much worse on other governance dimensions that their overall governance was weak. I also examine this issue in Section 4. But in order to do this, I first need to describe what constitutes good governance for financial firms. I do this in the next section in the context of the United States.

III. What is a Well-Governed Financial Firm?

Descriptions in the media of what appear to be egregious governance failures at financial institutions have heightened the impression that governance failures play a large part in the financial crisis. For example, boards are being blamed for what appear to be excessive pay packages that executives of financial firms received even while their firms were failing or being bailed out by the government. Morgenson (2009) reports that executives at seven major financial institutions that were in distress received $464 million in performance pay since 2005, while reporting losses of $107 billion since 2007. However, it is important to keep in mind that the media stories often describe individual cases, not the industry as a whole. To understand the role governance plays in the financial crisis, it is important to get a broader perspective of potential governance problems in the financial industry. Certainly any broad-based policy reform should not be based on the consideration of isolated cases.

Ideally, the academic governance literature would provide guidance on the question of whether financial institutions are well governed or not. However, because of the special nature of financial services, most academic papers exclude firms in the financial services from their data and focus on the governance of nonfinancial firms. Thus, to obtain a picture of the state of governance in the financial service industry, it is useful to directly examine some data on board characteristics and executive compensation. Before turning to the data, however, it is important to have a picture of what the board of a well-governed financial firm should look like.

Boards of financial firms have the same legal responsibilities as boards of nonfinancial firms, i.e., the duty of care and loyalty. In addition, publicly traded financial firms have to abide by SOX and exchange listing requirements. However, understanding what constitutes an effective governance structure for a financial firm is complicated by several factors. First, as Adams (2010) and Adams and Mehran (2003) describe, boards of financial firms may face more pressure to satisfy non-shareholder stakeholders than boards of nonfinancial firms. Regulators, for example, expect boards to act to ensure the safety and soundness of the financial institution, an objective that may not necessarily be in shareholders' best interest. Consistent with the idea that regulators and owners' interests may diverge, Laeven and Levine (2009) find in a cross-country analysis that the impact of regulation on bank risk taking depends on a bank's ownership structure. Adams and Mehran (2003, endnote 6) provide some examples of additional duties regulators impose on bank boards for the purpose of ensuring soundness, which include the adoption of real estate appraisal and evaluation policies (Federal Reserve Board Commercial Bank Examination Manual) and the annual approval of bank risk management policies (Federal Reserve Board Trading Activities Manual).

Second, financial firms are regulated by several different regulators. Investment banks are regulated by the Securities and Exchange Commission (SEC). Until recently, thrifts were regulated by the Office of Thrift Supervision.7 All banks with FDIC-insured deposits are subject to FDIC regulations. Because banks can choose to have a national or state charter and whether or not to be a member of the Federal Reserve, they effectively choose their regulatory authority. National banks are regulated by the OCC, while state banks are regulated either by the Federal Reserve or the FDIC. The presence of a regulator raises the question of whether regulatory scrutiny complements or substitutes for board-level governance. There is as yet no satisfactory answer to this question. Furthermore, it is not known whether regulators differ in the intensity with which they scrutinize the boards of the firms they examine. Some have argued that regulators may engage in a race to the bottom in order to attract banks with lax restrictions (see, e.g., Rosen 2003, 2005 and Whalen, 2002). Thus, it is possible that some regulators are more lenient in evaluating bank board behavior than others. For example, even though Federal Reserve Banks in theory penalize directors for poor attendance behavior, Adams and Ferreira (2011) find that the attendance behavior of directors of bank holding companies (BHCs) at board meetings is worse than in nonfinancial firms. Heterogeneity of regulators suggests that board-level governance may not be the same across all types of financial firms and the governance of each type of firm may need to be considered separately.

Third, the financial services industry underwent many recent changes that are likely to impact board governance. The banking industry underwent an intense period of consolidation in the 1990s through M&A activity. M&A activity affects board structure in several ways. First, it is common to add directors of target firms to the board of the acquirer in friendly acquisitions, as most banking M&As are. Adams and Mehran (2011) show that M&A activity leads to an increase in BHC board size. Second, M&A activity may have disciplining effects even if it is friendly. Thus, an active market for corporate control may improve board effectiveness. In the 1990s, banks were also increasingly allowed to engage in investment banking activities (culminating with the passage of the Gramm–Leach–Bliley Act in 1999). This created competition for investment banks, which may have put pressure on their boards. Consistent with this idea, Altınkılıç et al. (2007) find that investment banks do not appear to be ineffectively governed during the 1990–2003 period.

These three factors notwithstanding; because their duties to shareholders are the same, the literature generally argues that the same standards should apply to boards of financial firms as to the boards of nonfinancial firms. The three most commonly studied features of board structure are board independence, board size, and the number of other directorships directors hold. The literature generally argues that boards that are more independent, i.e., they contain more directors without social or business connections to management, should be more effective (see Adams et al. 2010 for a survey of the board literature). Smaller boards should be more effective because decision-making costs are lower in smaller groups. Because directors may become too busy when they hold more outside directorships, the literature argues that boards are more effective when directors hold fewer outside directorships.

Other features of boards that may be important for good governance are the attendance behavior of directors and the fraction of female directors on the board. The OECD's report on governance lessons from the crisis states that ‘boards' access to information is key’ (Kirkpatrick 2009, p. 23). In order to obtain information, directors need to attend board meetings.

Although it is not clear that having more female directors necessarily improves firm performance, Adams and Ferreira (2009) provide evidence that boardroom gender diversity may improve several important aspects of board behavior, for example, director attendance at board meetings. Moreover, Harriet Harman, UK Labor Party's number 2, famously argued that the financial crisis would have been less severe if Lehman Brothers had been Lehman Sisters (Morris 2009). Similarly, EU commissioner Michel Barnier suggested that having more women on bank boards would end the kind of ‘group-think’ that exacerbated the crisis (Treanor 2011).

It is less clear from the literature what effective CEO and director compensation should look like. To align their incentives with those of shareholders, CEOs and directors should receive a certain amount of performance-based pay in the form of equity. In addition, holding performance pay constant, total compensation should increase as risk increases. However, equity incentives may induce managers to take excessive risks. In addition, poorly governed firms may be more likely to overpay their directors. Thus, it is not always clear whether a given compensation contract is effective or not. Using an industry study, Philippon and Reshef (2009) argue that bankers were overpaid during the mid-1990s to 2006; however, they do not control for individual firm characteristics, such as size or risk, that may influence compensation packages.

Because in theory the same general governance standards apply to boards of both financial and nonfinancial firms, the governance of nonfinancial firms may serve as a useful benchmark for evaluating the governance of financial firms. Because the academic literature has studied the governance of nonfinancial firms in more depth, the governance of nonfinancial firms may be closer to an ‘ideal’ governance structure simply because of greater scrutiny and external pressure. If governance failures at financial firms are partly to blame for the financial crisis, we might therefore expect that financial firms have worse governance in terms of board characteristics and incentives than nonfinancial firms. In a sample of data on 35 BHCs ending in 1999, Adams and Mehran (2003) find that BHCs have larger boards, more independent directors, and lower performance-based pay for CEOs than nonfinancial firms. Thus, in their sample banks could be considered to be better governed than nonfinancial firms in some aspects (independence) but worse in other aspects (board size and performance pay). Because their sample predates the Gramm–Leach Bliley Act of 1999 which may have influenced governance structures in the industry, it is worthwhile making a similar comparison in a larger sample of more recent data, as I do in the next sections.

I choose eight governance characteristics (independence, board size, number of directorships, fraction of directors with attendance problems at board meetings, fraction of female directors, total CEO compensation, fraction equity-based pay for the CEO, and director compensation) and compare them across financial institutions and nonfinancial firms in a large sample of data. The arguments I describe above are frequently used to predict how these characteristics are related to governance quality. But, academics do not all agree on these predictions. Thus, for the sake of the argument, I categorize increases (or decreases) in a governance characteristic as a governance quality improvement if there is a regulation or regulatory recommendation promoting increases (or decreases) in that particular characteristic.

For example, SOX and the NYSE and Nasdaq exchange listing standards promote greater board independence. The UK's Walker Review (2009) suggests that bank board size should be reduced. In France and Germany, there are limits on the number of directorships directors can hold. The Federal Reserve System emphasizes the importance of attending board meetings in the bank director training book published by the Federal Reserve Bank of Kansas City (2010). This book lists attendance at board meetings as the second most important managerial function of bank directors (p. 50). Numerous countries are implementing boardroom gender quotas (see Deloitte 2011). Finally, the provisions in the 2010 Dodd–Frank Act concerning shareholders' right to a ‘say-on-pay’ can be interpreted as advocating lower executive pay and less incentive pay. For example, Protess (2011) writes that ‘Lawmakers are hoping to discourage companies from awarding lucrative packages that encourage risky behavior.’ Although I am unaware of any regulation specifically focusing on director compensation, by analogy with CEO compensation one can argue that decreasing director pay is seen to be a sign of better governance.

Table 1 summarizes how governance quality increases according to these regulations. Using the predictions in Table 1, I can also construct governance indices which enable me to compare several features of governance at a time. For each characteristic I consider, I define a dummy variable which is equal to 1 in a firm year if the firm has better than the median quality governance for that characteristic. As I describe in columns V and VI, I define a simple governance index, Governance index-7, to be the sum of the first seven dummy variables. Because compensation data are not as complete as other data in my sample, I also construct Governance index-5 which leaves out all compensation-based dummy variables

Table 1. Definition of ‘good governance’
Governance characteristicRegulatory policy concerning characteristicReference for column IIAccording to policy in column II, governance improves as characteristic …Governance index-7Governance index-5
IIIIIIIVVVI
  1. This table describes how eight governance characteristics affect governance quality according to various regulatory movements concerning governance. For each characteristic in column I, column II describes the regulatory policy or recommendation discussing that characteristic. Column III provides a reference for the policy or recommendation. I provide no reference for SOX or the exchange listing standards, as these are well documented in numerous sources. I provide no information for director compensation in column II, as I am unaware of policy specifically concerning outside director compensation. Column IV describes how the governance characteristic is supposed to improve governance quality. In columns V and VI, I describe how I use the information in column IV to construct governance indices. To construct Governance index-7, I assign each characteristic in each firm year a dummy variable which is equal to 1 if the characteristic is in the direction of good governance according to column IV. I use median levels of the characteristic to define the cutoffs for good governance. Governance index-7 is simply the sum of seven dummy variables. Governance index-5 is defined similarly, except that I exclude CEO compensation variables because of missing data. I do not consider director compensation in the construction of the governance indices because it is only available as an average number for 2006 and 2007 in Execucomp.
Board independenceSOX, NYSE, and Nasdaq listing standards increases1 if above sample median, otherwise 01 if above sample median, otherwise 0
Board sizeUK's Walker ReviewWalker (2009)decreases1 if below sample median, otherwise 01 if below sample median, otherwise 0
# DirectorshipsFrance's New Economic Regulation of 2001, German company law, Article 100

http://www.practicallaw.com/5-107-0184?q=&qp=&qo=&qe=,

http://www.aktiengesetz.de/

decreases1 if below sample median, otherwise 01 if below sample median, otherwise 0
Fraction attendance problemsFederal Reserve SystemFederal Reserve Bank of Kansas City (2010)decreases1 if below sample median, otherwise 01 if below sample median, otherwise 0
Fraction femaleNumerous country-level quotas on boardroom gender diversity, also proposed quota on bank boards by Michel Barnier, EU Internal Markets CommissionerDeloitte (2011), Treanor (2011)increases1 if above sample median, otherwise 01 if above sample median, otherwise 0
Total CEO compensationDodd-Frank Act of 2010Protess (2011)decreases1 if below sample median, otherwise 0
Fraction equity-based pay CEODodd-Frank Act of 2010Protess (2011)decreases1 if below sample median, otherwise 0
(Nonexecutive) Director compensationdecreases (by analogy with CEO compensation)

IV. The Governance of Financial Firms – What are the Facts?

In this section, I examine whether potential governance failures of financial firms in the United States can be attributed to the fact that they were noncompliant with SOX and the listing standards. Next I examine whether financial firms complied with some aspects of governance standards but were so much worse on other governance dimensions that their overall governance was weak. But first I describe the data I use.

A. Data

To compare board characteristics and incentives in financial firms and nonfinancial firms, I use the Riskmetrics director database from 1996 to 2007.8 This is an unbalanced panel of director-level data for Standard & Poor's (S&P) 500, S&P MidCaps, and S&P SmallCap firms. It contains information on directors from company proxy statements or annual reports, such as whether the director is classified as independent and the number of other directorships each director holds. I merge these data to Compustat to obtain SIC codes, which I use to identify financial firms and banks. I obtain data on financial characteristics and CEO and director compensation from ExecuComp.

Using these data, I construct a firm-level data set containing 18,839 firm-year level observations. I define fraction of directors with attendance problems to be the fraction of directors who attended fewer than 75% of meetings they were supposed to attend in the prior fiscal year. Total CEO compensation is Execucomp item TDC1. Fraction equity-based pay CEO is 1 − (salary + bonus)/TDC1. All compensation numbers are adjusted to 2007 dollars using the CPI-U. Director compensation is only available for 2006 and 2007 because of changes in reporting requirements. Return on assets, ROA, is Execucomp item ROA. Tobin's Q is (book value of assets − book value of equity + market value of equity)/book value of assets [(assets − commeq + mktval)/assets]. Volatility is the standard deviation of prior 60-month stock returns. All other variables are described in Table 2. I define board size, assets, total CEO compensation, and director compensation to be missing if they are less than 3, 0, or nonpositive, respectively.

Table 2. Summary statistics
VariableObsMeanStd. Dev.MinMax
  1. The data for panels A–C consist of an unbalanced panel of 18,839 firm-year level observations for S&P 1500 firms for the period 1996–2007 which were in the Riskmetrics Director Database. Data on assets and SIC codes are from Compustat. Ln(Assets) is the natural logarithm of assets. Data for return on assets (Execucomp item: ROA), Tobin's Q, volatility (Execucomp item: bs_volatility), and CEO and director compensation are from Execucomp. Tobin's Q is defined to be (assets − commeq + mktval)/assets. Volatility is prior 60-month stock return volatility. SIC codes were used to classify firms into financial firms, banks, and nonfinancial firms. Financial firm is a dummy equal to 1 if the firm is a financial firm. Bank is a dummy variable equal to 1 if the sample firm is a bank. Riskmetrics classifies directors as independent if they have no business relationship with the firm, are not related or interlocked with management, and are not current or former employees. Board independence is the number of independent directors on the board divided by board size. Board size is set to missing if it is less than 3. # Directorships is the average number of directorships in other for-profit companies per director. Fraction attendance problem is the fraction of directors who attended fewer than 75% of the meetings they were supposed to attend during the previous fiscal year. Fraction female is the fraction of female directors. Total CEO compensation (Execucomp item: TDC1) is the sum of salary, bonus, other annual, total value of restricted stock granted, total value of stock options granted (using Black-Scholes), long-term incentive payouts, and all other total. Total CEO compensation is set to missing if it is 0. Fraction equity-based pay CEO is 1 − (Salary + Bonus)/Total CEO Compensation. Director compensation is only available for the years 2006 and 2007 and is the average compensation for (generally nonexecutive) directors. Total CEO compensation is measured in millions. Director compensation is measured in thousands. All compensation figures have been converted to 2007 dollars using the CPI-U. Assets are measured in millions. The governance indices are defined in Table 1. Observations vary because of missing data.
Panel A: firm characteristics
Assets18,12413,356.85068,035.9506.2682,187,631
Ln(Assets)18,1247.6411.6791.83514.598
ROA16,3153.37616.449−587.97370.325
Tobin's Q16,1551.9711.7310.40477.635
Volatility14,6640.4090.2000.1083.284
Financial firm18,5540.1460.35301
Bank18,5320.0610.24001
Panel B: firm-level governance characteristics
Board independence18,8390.6520.18201
Board size18,8379.5292.910339
# Directorships15,8110.8170.62004.786
Fraction attendance problem18,8390.0190.05100.545
Fraction female18,8210.0900.09001
Total CEO compensation15,5994.86220.3830.0002193.705
Fraction equity-based pay CEO15,5990.5270.28101
Director compensation1871174.866129.1219.8702445.206
Panel C: firm-level governance indices
Governance index-713,1783.8821.41407
Governance index-515,8102.9381.04205
Post-reform Governance index-761693.8171.38307

Financial firms comprise 14.34% of the sample (2702 observations). Of the financial firm observations, 1136 observations belong to banks. Table 2 shows summary statistics for firm characteristics and the eight governance characteristics. Table 2 also shows summary statistics for Governance index-7 and Governance index-5. The number of observations varies across rows because of missing data.

All of the firms in this database are publicly traded firms and all banks are BHCs, thus these data do not allow me to shed any light on the governance of private financial firms. The number of BHCs represented in the sample varies over time from a low of 86 to a high of 105. The number of nonbank financial firms (NBFFs) varies from 106 to 158. Although the number of banks in the sample is small relative to the banking industry, the banks in this sample represent a large fraction of industry assets. For example, according to the FDIC, in 2007 there were 7282 FDIC-insured commercial banks in the United States with a total of $11,176 billion in assets. In 2007, there are 93 banks in my sample with assets comprising 49.86% of total commercial bank assets ($5,572,369 million). Thus, although the comparisons I make and conclusions I draw need not apply to all banks, they are relevant for understanding potential governance failures at banks that are likely to matter the most for the crisis.

B. Did financial firms comply with SOX and the listing standards?

If financial firms in the United States did not comply with SOX and the listing standards, then this would be a simple explanation why governance may have contributed to the financial crisis. Although a large literature examines the implications of SOX and the listing standards for nonfinancial firms, less is known about the effects of these reforms on financial firms. A full examination of compliance by financial firms is beyond the scope of this paper. However, I can provide some suggestive evidence on this issue by examining changes in one governance characteristic emphasized in SOX and the listing standards, which I label US reforms, namely board independence.

In Table 3, I compare board independence in the years after 2002 to board independence prior by regressing board independence on a dummy variable, Post-US Reform, which is equal to 1 for all years after and including 2002. In column I, I show the regression for the full sample. In columns II–IV, I divide the sample into nonfinancial firms, banks, and nonbank financials. The regressions in columns V–VII replicate those in II–IV, except with firm fixed effects. The coefficients on Post-US Reform are positive and significant at greater than the 1% level across all columns.

Table 3. Board independence post-governance reform in the US and precrisis
 Board independence
IIIIIIIVVVIVII
  1. The table shows OLS regressions of board independence on a dummy variable (Post-US Reform) which is equal to 1 for all years after and including 2002. The data set in consist of Riskmetrics data from 1996 to 2007. The columns vary by the type of firms and the inclusion of firm fixed effects (columns V–VII). The sample in column I consists of the full sample of Riskmetrics data as described in Table 2. The sample in columns II and IV consists of nonfinancial firms. The sample in columns III and VI consists of banks. The sample in columns III and VII consists of nonbank financial firms. The classification of firms into categories is described in Table 2. Observations vary because of missing data. Absolute values of t-statistics are in brackets. Standard errors are heteroskedasticity corrected in columns V–VII. Asterisks indicate significance at 0.01 (***), 0.05 (**), and 0.10 (*) levels.
Post-US Reform0.1050.1070.0360.1280.0940.0290.115
[41.15][38.39][4.34][14.66][27.45][2.67][8.17]
Constant0.6040.6030.6800.5640.6090.6830.570
[349.57][317.79][119.40][91.70][384.74][130.32][81.92]
Sample typeFull sampleNonfinancial firmsBanksNonbank financialsNonfinancial firmsBanksNonbank financials
Firm fixed effects?NoNoNoNoYesYesYes
Obs.18,83915,8521134156815,85211341568
R-squared0.0820.0850.0160.1210.1560.0220.182

The results for nonfinancial firms are consistent with the findings from prior literature that the US reforms led to an increase in board independence. The results for financial firms suggest that the same is true for financial firms. The magnitudes of the coefficients on Post-US Reform are always highest for NBFFs and are the lowest for banks. This indicates that NBFFs increased board independence the most and banks the least following the implementation of SOX and the listing standards. However, the magnitudes of the coefficients on the constant terms indicate that banks had the highest level of board independence prior to the US reforms which explains their smaller average response to the reforms.

These results are suggestive evidence that financial firms complied with the US reforms emphasizing board independence. However, it is possible that financial firms performed poorly along other governance dimensions leading their overall governance quality to be weak. To examine this hypothesis, I compare the governance of financial firms to that of nonfinancial firms in the next section.

C. Comparing the governance of financial firms to that of nonfinancial firms

Because of the reasons outlined in Section 3, I compare the governance characteristics of banks and nonbank financials to those of nonfinancial firms separately. Table 4 provides comparisons for banks and nonfinancial firms. Table 5 provides the same comparisons for NBFFs. Columns I–VIII of Table 4 show comparisons of means of the eight governance characteristics from Table 1 (independence, board size, number of directorships, fraction of directors with attendance problems at board meetings, fraction of female directors, total CEO compensation, fraction equity-based pay for the CEO, and director compensation) for banks and nonfinancial firms. Columns I–VIII are for the entire sample from 1996 to 2007. Columns IX–XVI repeat the comparisons for 2007 data only. Observations vary because of missing data. The coefficient on ‘Bank’ (‘NBFF’) is the coefficient on a ‘Bank’ (‘NBFF’) dummy variable in an ordinary least squares regression with the respective governance characteristic as the dependent variable. Results vary across panels because of the inclusions of different control variables in the regressions. In panel A, regressions contain no control variables. In panel B, regressions contain Ln(Assets) as a proxy for firm size and standard errors are clustered at the firm level. In panel C, I include Ln(Assets), ROA, Tobin's Q, and year dummies, and cluster the standard errors at the firm level. The regressions in columns I–VIII include volatility as well. I omit volatility from the regressions in columns IX–XVI because of missing data. Asterisks indicate statistical significance, with ***, **, and * indicating significance at the 1, 5, and 10% levels, respectively.

Table 4. Comparison of selected governance characteristics of banks to nonfinancial firms
 Entire sample 1996–2007
Board independenceBoard size# DirectorshipsFraction attendance problemsFraction femaleTotal CEO compensationFraction equity-based pay CEODirector compensation
IIIIIIIVVVIVIIVIII
Panel A: OLS regressions of governance characteristics on the bank dummy in sample of banks and nonfinancial firms
Bank0.0454.631−0.3070.0060.0070.0480.001−46.579
[8.05][56.96][14.76][3.72][2.50][0.07][0.11][3.79]
Obs.169861698414276169861696914408144081666
Score10101NANA1
Total governance score for banks: 4/6
Panel B: OLS regressions of governance characteristics on the bank dummy and Ln(Assets) – standard errors clustered at firm level
Bank−0.0132.319−0.7770.006−0.037−5.478−0.131−107.720
[1.13][7.41][18.21][2.39][5.84][7.82][8.74][7.90]
Obs.166141661213954166141659714408144081666
ScoreNA0100111
Total governance score for banks: 4/7
Panel C: OLS regressions of governance characteristics on the bank dummy, Ln(Assets), ROA, Tobins' Q, volatility (in left panel only), year dummies – standard errors clustered at firm level
Bank−0.0151.917−0.7830.003−0.041−4.431−0.091−75.620
[1.20][5.79][15.84][1.16][5.49][6.67][6.42][4.03]
Obs.13433134331107913433134311308013080966
ScoreNA01NA0111
Total governance score for banks: 4/6
 2007 data only
Board independenceBoard size# DirectorshipsFraction attendance problemsFraction femaleTotal CEO compensationFraction equity-based pay CEODirector compensation
IXXXIXIIXIIIXIVXVXVI
  1. The table compares selected governance characteristics for banks and nonfinancial firms by regressing governance characteristics on a dummy variable (‘Bank’) equal to 1 if the sample firm is a bank. The panels vary by the control variables included in the regressions. The sample consists of an unbalanced panel of 16,986 firm-year level observations on banks and nonfinancial firms for the period 1996–2007 which were in the Riskmetrics Director Database. The sample and governance characteristics are described in Table 2. Volatility is excluded from the regressions in panel C using 2007 data only because of missing data. Observations vary because of missing data. The coefficients on the control variables and the constant are omitted for the sake of brevity. Absolute values of t-statistics are in brackets. Standard errors are clustered at the firm level in panels B and C. Asterisks indicate significance at 0.01 (***), 0.05 (**), and 0.10 (*) levels. For each characteristic for which the coefficient on ‘Bank’ is significant at greater than the 10% level, ‘Score’ is defined to be 1 if banks appear better governed according to column IV of Table 1, otherwise ‘Score’ is defined to be 0. If the coefficient on ‘Bank’ is insignificant, ‘Score’ is equal to NA. For each sample and each specification, ‘Total governance score for banks’ shows the sum of ‘Score’ divided by the number of significantly different coefficients on ‘Bank’ for that sample and specification. Thus, ‘Total governance score for banks’ is an index measuring the relative strength of bank governance as compared with nonfinancial firm governance according to Table 1.
Panel A: OLS regressions of governance characteristics on the bank dummy in sample of banks and nonfinancial firms
Bank−0.0083.133−0.484−0.0040.008−0.868−0.076−65.561
[0.64][13.04][8.57][0.97][0.73][0.75][2.16][3.17]
Obs.12681268126812681268702702698
ScoreNA01NANANA11
Total governance score for banks: 2/4
Panel B: OLS regressions of governance characteristics on the bank dummy and Ln(Assets) – standard errors clustered at firm level
Bank−0.0361.419−0.797−0.003−0.027−6.845−0.204−127.855
[3.05][5.10][18.98][1.02][2.94][8.66][5.68][9.02]
Obs.12591259125912591259702702698
Score001NA0111
Total governance score for banks: 4/7
Panel C: OLS regressions of governance characteristics on the bank dummy, Ln(Assets), ROA, Tobins' Q, volatility (in left panel only), year dummies – standard errors clustered at firm level
Bank−0.0391.126−0.744−0.004−0.034−6.492−0.189−116.560
[2.36][2.85][11.27][1.31][2.40][8.10][5.26][8.13]
Obs.701701701701701701701697
Score001NA0111
Total governance score for banks: 4/7
Table 5. Comparison of selected governance characteristics of nonbank financial firms to nonfinancial firms
 Entire sample 1996–2007
Board independenceBoard size# DirectorshipsFraction attendance problemsFraction femaleTotal CEO compensationFraction equity-based pay CEODirector compensation
IIIIIIIVVVIVIIVIII
Panel A: OLS regressions of governance characteristics on the nonbank financial dummy in sample of nonbank financial firms and nonfinancial firms
NBFF−0.0251.210−0.0300.0030.0032.6640.0304.858
[5.13][17.98][1.73][1.85][1.15][4.21][3.46][0.51]
Obs.174201741814679174201740314679146791747
Score0010NA00NA
Total governance score for NBFFs: 1/6
Panel B: OLS regressions of governance characteristics on the nonbank financial dummy and Ln(assets) – standard errors clustered at firm level
NBFF−0.061−0.244−0.3090.002−0.025−1.262−0.060−33.790
[5.53][1.37][7.75][1.16][4.72][1.88][3.64][3.59]
Obs.170431704114350170431702614679146791747
Score0NA1NA0111
Total governance score for NBFFs: 4/6
Panel C: OLS regressions of governance characteristics on the nonbank financial dummy, Ln(assets), ROA, Tobins' Q, volatility (in left panel only), year dummies – standard errors clustered at firm level
NBFF−0.063−0.259−0.3420.004−0.029−1.022−0.052−27.421
[4.95][1.26][7.35][1.40][4.67][1.50][3.00][2.59]
Obs.13654136541129213654136521327413274998
Score0NA1NA0NA11
Total governance score for NBFFs: 3/5
 2007 data only
Board independenceBoard size# DirectorshipsFraction attendance problemsFraction femaleTotal CEO compensationFraction equity-based pay CEODirector compensation
IXXXIXIIXIIIXIVXVXVI
  1. The table compares selected governance characteristics for nonbank financial firms and nonfinancial firms by regressing governance characteristics on a dummy variable (NBFF) equal to 1 if the sample firm is a financial firm that is not a bank. The panels vary by the control variables included in the regressions. The sample consists of an unbalanced panel of firm-year level observations on financial institutions and nonfinancial firms for the period 1996–2007 which were in the Riskmetrics Director Database. The sample and governance characteristics are described in Table 2. Volatility is excluded from the regressions in panel C using 2007 data only because of missing data. Observations vary because of missing data. The coefficients on the control variables and the constant are omitted for the sake of brevity. Absolute values of t-statistics are in brackets. Standard errors are clustered at the firm level in panels B and C. Asterisks indicate significance at 0.01 (***), 0.05 (**), and 0.10 (*) levels. For each characteristic for which the coefficient on NBFF is significant at greater than the 10% level, ‘Score’ is defined to be 1 if nonbank financials appear better governed according to column IV of Table 1, otherwise ‘Score’ is defined to be 0. If the coefficient on NBFF is insignificant, ‘Score’ is equal to NA. For each sample and each specification, ‘Total governance score for NBFFs’ shows the sum of ‘Score’ divided by the number of significantly different coefficients on NBFF for that sample and specification. Thus, ‘Total governance score for NBFFs’ is an index measuring the relative strength of nonbank financial firm governance as compared with nonfinancial firm governance according to Table 1.
Panel A: OLS regressions of governance characteristics on the nonbank financial dummy in sample of nonbank financial firms and nonfinancial firms
NBFF−0.0310.932−0.081−0.005−0.0050.93−0.0382.208
[3.24][4.91][1.81][1.70][0.53][1.06][1.47][0.15]
Obs.13351335133513351335751751747
Score0011NANANANA
Total governance score for NBFFs: 2/4
Panel B: OLS regressions of governance characteristics on the nonbank financial dummy and Ln(assets) – standard errors clustered at firm level
NBFF−0.053−0.166−0.269−0.004−0.028−2.521−0.105−32.723
[5.56][0.83][6.00][2.30][3.56][3.06][3.59][2.79]
Obs.13241324132413241324751751747
Score0NA110111
Total governance score for NBFFs: 5/7
Panel C: OLS regressions of governance characteristics on the nonbank financial dummy, Ln(assets), ROA, Tobins' Q, volatility (in left panel only), year dummies – standard errors clustered at firm level
NBFF−0.051−0.296−0.213−0.002−0.025−2.304−0.096−25.964
[4.32][1.11][3.55][0.89][2.38][2.80][3.31][2.28]
Obs.750750750750750750750746
Score0NA1NA0111
Total governance score for NBFFs: 4/6

From columns I–VIII of Table 1, panel A, we see that banks have on average more independent boards, larger boards, fewer outside directorships, greater attendance problems, a greater fraction of female directors, and lower director compensation than nonfinancial firms. While board size is larger in banks, Adams and Mehran (2011) do not find that larger bank board size has detrimental effects on shareholder value. Thus, on average banks do not appear to be worse governed than nonfinancial firms, except possibly in terms of attendance behavior of directors. The latter finding is consistent with Adams and Ferreira (2011).

In contrast to expectations, banks have on average more boardroom diversity than nonfinancial firms. However, it is possible that the results are skewed because banks are generally much larger than nonfinancial firms in terms of assets. Thus, in panel B, I perform the same comparisons after controlling for firm size, as proxied by the natural logarithm of the book value of assets.9 The picture looks slightly different now. Controlling for size, bank boards are still larger, with fewer outside directorships and greater attendance problems. However, they now have a smaller fraction of female directors. But they also have less total CEO compensation, less incentive pay for the CEO, and less director compensation than nonfinancial firms. These results hold even in the 2007 data. Clearly, firm size is an important factor influencing governance characteristics, consistent with findings in Boone et al. (2007), Coles et al. (2008), Lehn et al. (2009), and Linck et al. (2008). For example, the coefficient on ‘Bank’ in the board size comparison decreases by roughly 50% in column II after controlling for firm size and the compensation numbers decrease significantly. These results are robust to controlling for additional firm characteristics in panel C.

Based on these comparisons, it would be difficult to argue that banks are clearly poorly governed. In particular, much of the media attention focuses on ‘excessive’ total pay and performance pay in financial firms, yet, on average, bank CEOs earn less and have less performance pay than CEOs of nonfinancial firms of similar size. Because banks perform worse on some dimensions of governance (according to Table 1) and better on other dimensions than nonfinancial firms, I summarize their overall governance performance relative to nonfinancial firms using a governance score. For each characteristic for which the coefficient on ‘Bank’ is significant at greater than the 10% level, I define the governance score to be 1 if banks appear better governed according to column IV of Table 1, otherwise the score is defined to be 0. For each sample and each panel in Table 4, ‘Total governance score for banks’ shows the sum of the scores divided by the number of significantly different coefficients on ‘Bank’ for that sample and panel.

The ‘Total governance score for banks’ varies from a low of 2/4 for panel A, 2007 data, to a high of 4/6 in panels A and C, 1996–2007 data. But in all cases, this score indicates that banks are generally equally, if not better, governed than nonfinancial firms. Although this measure is ad hoc because it gives equal weight to each dimension of governance, it illustrates that understanding potential governance failures at banks is not straightforward.

The picture looks slightly different when we compare NBFFs to nonfinancial firms in Table 5. Comparing panel A to panel B for 1996–2007 data, it is clear that firm size again has a significant effect on governance characteristics. After controlling for firm size, NBFFs have less independent boards, fewer outside directorships, a lower fraction of female directors, lower total CEO compensation, lower performance pay, and lower director compensation than nonfinancial firms. Moreover, the governance score for NBFF changes from 1/6 to 4/6 after controlling for firm size. Similar to the scores for banks, the governance score for NBFFs indicates for all but panel A, 1996–2007 data that NBFFs are at least as well governed as nonfinancial firms.

In Table 6, I perform a similar comparison as in Tables 4 and 5, except now I use Governance index-7 and Governance index-5 as dependent variables in my regressions. I also define a post-reform Governance index-7 that uses only data from 2003 to 2007 to define the medians in generating the index. Panel A shows results of regressions of the indices on ‘Bank’ and control variables. Panel B shows the same regressions for NBFFs.

Table 6. Comparisons of governance indices for financial institutions and nonfinancial firms
 Dependent variable: Governance index-7Dependent variable: Governance index-5Dependent variable: Post-reform Governance index-7
IIIIIIIVVVIVIIVIIIIX
  1. The table compares governance indices for financial firms and nonfinancial firms by regressing governance indices on a dummy variable bank (or NBFF) equal to 1 if the sample firm is a bank (or a financial firm that is not a bank). In panel A, the main explanatory variable is bank, and nonbank financial firms are excluded from the sample. In panel B, the main explanatory variable is NBFF, and banks are excluded from the sample. The columns vary by the control variables included in the regressions. The sample consists of an unbalanced panel of firm-year level observations on financial firms and nonfinancial firms for the period 1996–2007 which were in the Riskmetrics Director Database. The sample and control variables are described in Table 2. Governance indices are defined as in Table 1. Post-reform Governance index-7 is defined in the same manner as Governance index-7 (see Table 1), except that the median levels of governance characteristics are defined only for data from 2003 to 2007. Observations vary because of missing data. The regressions in columns III, VI, and IX include year dummies. Standard errors are clustered at the firm level in all regressions except those in columns I, IV, and VII. Absolute values of t-statistics are in brackets. Asterisks indicate significance at 0.01 (***), 0.05 (**), and 0.10 (*) levels.
Panel A: Banks versus nonfinancial firms
Bank−0.1740.8750.819−0.2000.2050.2800.0571.1061.001
[3.33][8.80][8.00][5.75][2.77][3.39][0.76][10.98][8.96]
Ln(Assets) −0.426−0.491 −0.162−0.199 −0.454−0.491
 [32.53][32.78] [16.51][16.50] [31.62][28.84]
ROA  -2.519e-04  0.001  0.003
  [0.28]  [1.40]  [1.04]
Tobins' Q  −0.064  0.007  −0.120
  [4.64]  [0.97]  [4.59]
Volatility  −0.904  −0.113  −0.496
  [7.40]  [1.21]  [3.57]
Constant3.9327.1217.8322.9784.1764.1173.8477.3307.872
[297.40][72.05][52.94][332.52][56.74][36.96][200.97][65.62][44.85]
Obs.121391213910817142751395311079561256124888
R-squared0.0010.2090.2530.0020.0560.126<0.0010.250.263
Panel B: Nonbank financial firms versus nonfinancial firms
NBFF−0.4990.1950.155−0.276−0.019−0.033−0.3670.3420.259
[10.94][2.16][1.64][9.31][0.31][0.45][5.97][3.69][2.63]
Ln(Assets) −0.400−0.462 −0.153−0.187 −0.425−0.464
 [30.92][30.74] [15.95][15.65] [29.06][27.14]
ROA  -4.12E-04  0.001  0.001
  [0.47]  [1.25]  [0.54]
Tobins' Q  −0.066  0.005  −0.120
  [4.83]  [0.67]  [4.57]
Volatility  −0.805  −0.078  −0.431
  [6.69]  [0.84]  [3.21]
Constant3.9326.9287.5862.9784.1084.0363.8477.1057.652
[298.14][70.73][53.15][331.24][57.09][34.78][201.77][62.32][43.57]
Obs.124031240311003146781434911292579757975024
R-squared0.010.2050.2480.0060.0570.1260.0060.2380.257

The results in Table 6 are similar to those in Tables 4 and 5. Superficially, banks and NBFF may appear worse governed than nonfinancial firms. For example, the coefficient on ‘Bank’ is significantly negative in columns I and IV and insignificant in column VII. However, after controlling for firm characteristics, banks appear to be better governed than nonfinancial firms regardless of the index used and NBFFs are at least as well governed.

Some conclusions one can draw from the discussion in this section are as follows: First, banks and NBFFs do not have exactly the same governance characteristics. Second, although executive pay in financial firms may appear larger than in nonfinancial firms, much of this difference is driven by differences in firm size. Of course, CEOs at some financial firms may have received what many consider to be unfair or excessive pay. However, on average, financial CEO pay does not seem to be excessive relative to CEO pay in nonfinancial firms. Moreover, the phenomenon of ‘excess pay’ for CEOs appears to be driven by NBFFs. Third, although financial firms score worse on some governance characteristics than nonfinancial firms, on average they appear to be better governed than nonfinancial firms. Nevertheless, it is possible that the differences in governance I document are still partly to blame for the problems facing financial firms. I turn to this issue in the next section.

A final and potentially very important observation is that director pay is significantly lower in banks, even without controlling for firm size. This fact is also pointed out in Adams and Ferreira (2011). From Table 4, panel A, column XVI, one can see that in 2007 average bank director compensation is lower than average nonfinancial director compensation by $65,561. Average director compensation in nonfinancials in 2007 is $193,931, so bank director compensation is 33.8% lower. This may have serious consequences for governance. If the pool of director candidates is the same for financial firms and nonfinancial firms, then the better qualified candidates may prefer to join the boards of firms that pay more. Of course, factors other than pay play a role in an individual's decision to join a board, but this finding raises the possibility that the pool of bank directors may be worse along some dimensions than the pool of nonfinancial firm directors.

V. Is Governance to Blame for the Problems at Banks?

In the previous section, I showed that the governance of financial firms differs from that of nonfinancial firms. Although financial firms do not appear worse governed, a natural question is whether these differences may still have contributed to the problems financial firms faced during the crisis. In general, this is not an easy question to answer. One way of trying to answer it is to compare the governance characteristics of financial firms that received bailout money from the Federal Government to those that did not. The Troubled Assets Relief Program (TARP) was designed to strengthen the financial system by enabling the government to purchase or insure up to $700 billion in troubled assets. Although the US treasury did not describe it as a bailout program (see http://www.financialstability.gov/roadtostability/capitalpurchaseprogram.html), many observers referred to it as such (see, e.g., the TARP entry on Wikipedia). Moreover, it is clear that recipients of TARP money faced some financial difficulties. Thus, a comparison of firms with and without TARP assistance may be illustrative. In addition, I can eliminate firms that failed or were acquired from my sample to ensure that I am comparing institutions that were essentially healthy that did not receive bailout funds to those that did.

As of April 10, 2009 only six nonbank financial firms received TARP assistance. Thus, I limit the comparison to banks. To determine which of the banks in my sample received TARP funds, I matched their names and locations to the names and locations in the list of TARP recipients maintained by the New York Times (http://projects.nytimes.com/creditcrisis/recipients/table). Of the 93 banks in my sample in 2007, 56 received bailout funds in either 2008 or beginning of 2009 (until April 10, 2009). To determine whether any of the remaining institutions failed or were acquired, I first merge my sample to data from Osiris gathered on April 8, 2009. This data set contains information about the current status of institutions. For the 16 institutions whose status I was unable to verify using Osiris, I checked institutional histories on the National Information Center's website,10 a website containing information on banks collected by the Federal Reserve System. Four institutions in my sample disappeared by 2009. Two were acquired (Wachovia and National City Corp), one closed (Downey Financial Corp), and one failed (Franklin Bank Corp). I eliminate these observations from my sample. In Table 7, I perform a similar comparison for 2007 as in Tables 4 and 5 after restricting my sample to banks. The coefficients on ‘Bailout’ measure differences in governance characteristics for banks that received TARP money in 2008 and 2009 to those that did not. Because the sample is so small, I do not also report the results after controlling for firm size. They are similar in sign, although less significant, except for director compensation that becomes significantly negative.

Table 7. Comparison of selected governance characteristics for sample banks receiving bailout money to surviving sample banks that did not
 2007 data only
Board independenceBoard size# DirectorshipsFraction attendance problemsFraction femaleTotal CEO compensationFraction equity-based pay CEODirector compensation
IIIIIIIVVVIVIIVIII
  1. The table shows a comparison of selected governance characteristics in 2007 between sample banks that received bailout money from the US government in 2008 and beginning of 2009 (up until April 10, 2009) and sample banks that survived until April, 2009 and did not receive bailout money. I define a bank to have received bailout money if it received funds from the US government under the Troubled Asset Relief Program (TARP). The underlying sample is described in Table 1. To determine which of the banks in my sample received TARP funds, I matched their names and locations to the names and locations in the list of TARP recipients maintained by the New York Times (http://projects.nytimes.com/creditcrisis/recipients/table). Of the 93 banks in my sample in 2007, 56 received bailout funds in either 2008 or beginning of 2009 (until April 10, 2009). To determine whether any of the remaining institutions failed or were acquired, I first merge my sample to data from Osiris gathered on April 8, 2009. This data set contains information about the current status of institutions as of beginning of 2009. For the 16 institutions whose status I was unable to verify using Osiris, I checked institutional histories on the National Information Center's website, a website containing information on banks collected by the Federal Reserve System. Four institutions in my sample that did not receive TARP funds disappeared by 2009. Two were acquired, one closed, and one failed. I eliminate these observations from my sample. Governance characteristics in columns I–VIII are defined as in Table 2. Bailout is a dummy variable equal to 1 if the sample firm received bailout money from the US government. Observations vary because of missing data. Absolute values of t-statistics are in brackets. Asterisks indicate significance at 0.01 (***), 0.05 (**), and 0.10 (*) levels.
Bailout0.0371.3400.2650.0060.0293.4490.163−3.699
[1.68][2.13][2.66][1.21][1.59][1.79][1.97][0.14]
Constant0.74511.3030.2460.0020.1063.1520.517131.295
[42.67][22.64][3.11][0.60][7.44][1.98][7.59][5.98]
Obs.8989898989444444
R-squared0.0310.0490.0750.0170.0280.0710.085<0.001

Table 7 indicates that banks with TARP funds have more independent boards, larger boards, more outside directorships, greater total CEO pay, and greater incentive pay for CEOs. Some of these results are consistent with the idea that TARP banks have worse governance. In particular, the fact that TARP banks had higher performance pay for CEOs is consistent with the idea that performance pay may have led executives of banks to take on too much risk. The coefficient on the number of directorships is also consistent with potentially worse governance, because taking on too many directorships can lead directors to become too unfocused. The fact that TARP banks have larger boards is also a potential indication of worse governance.

Perhaps the most surprising result, however, is the finding that TARP banks have boards that are more independent. What is going on here? There are several possible explanations. For example, it is possible that the governance of TARP banks is not much worse than that of non-TARP banks. However, given that a large part of the problems at banks was caused by securitization, another explanation is possible. An independent director, by definition, is a director who has not worked for the bank and has no business dealings with the bank. Because of potential conflicts of interests, independent directors are generally not employees of other financial firms. What this means is that independent directors are less likely to have an in-depth knowledge of the internal workings of the banks on whose boards they sit. They are also less likely to have the financial expertise to understand the complexity of the securitization processes banks were engaging in or to assess the associated risks banks were taking on. Thus, although board independence is generally seen to be a good thing, in the case of banks, greater independence may be a bad thing because a more independent board will not have sufficient expertise to monitor the actions of the CEO. This finding is also consistent with Guerrera and Larsen (2008) who describe that more than two-thirds of the directors at eight large US financial institutions did not have any significant recent experience in the banking industry and more than half had no financial service experience at all.

If we accept that independence may be a sign of poor governance, then the conclusion to be drawn from Table 7 is that it appears that TARP banks were indeed governed worse than non-TARP banks. However, because the messages from Tables 4 and 5 are mixed, i.e., banks and financial firms do not necessarily appear to be worse governed than nonfinancial firms, it is not obvious what the policy implications are. I discuss some lessons we can learn in the next section.

VI. What Lessons can we Learn?

Because it is the duty of the board to oversee management, and many bank and financial firm managers led their banks to the brink of failure, boards of financial firms clearly share some responsibility for the crisis. But the question is how much of the blame should they shoulder? Popular opinion, fueled by stories of what seems to be excessive executive compensation at financial firms, might argue: a lot. However, it is important to keep several facts in mind, particularly when considering potential policy implications.

First, few people predicted the financial crisis. Rushe (2008) describes, for example, how little attention was paid to Nouriel Roubini's prediction that problems in the subprime mortgage market would trigger a financial crisis. While the boards of financial firms should have better information than outsiders, directors are generally not experts on the economy. Thus, it seems unreasonable to expect that they should have been better able to predict the problems financial firms would face than academics, regulators, and financial analysts. Rodrik (2009) argues in his blog that blame should be spread more widely than bankers to the broader economics and policymaking community.

Second, financial firms are regulated. It is still not clear whether regulators substitute or complement board-level governance. However, it is possible that directors perceive a substitution effect. It is not hard to imagine that a bank director does not understand all risk implications of particular transactions but agrees to them anyhow, because he assumes that regulators would identify any potential problems.

Third, the data in this paper show that governance in financial firms is, on average, not obviously worse than in nonfinancial firms. While financial firm governance may appear worse in some dimensions, it appears better in others. Ex post, it is easy to argue that governance problems occurred, but ex ante it is not clear that boards of financial firms were doing anything much different from boards in other firms. Even the issue of executive compensation is not as clear cut as it appears at first. CEOs of NBFFs do earn significantly more than CEOs of nonfinancial firms, but this is no longer true once firm size is accounted for.

However, two particular findings suggest that banking firm governance may have been worse, but in ways that might not have been predicted ex ante. Banks receiving bailout money had boards that were more independent and bank directors earned significantly less compensation than their counterparts in nonfinancial firms. What this suggests is that board independence may not necessarily be beneficial for banks. Independent directors may not always have the expertise necessary to oversee complex banking firms.

The fact that bank directors earn so much less than their counterparties in nonfinancial firms raises the possibility that the pool of bank directors is different from the pool of directors of nonfinancial firms. Further research is needed to examine why director pay is so low in banking.11 However, regardless of whether low pay is a sign of a governance problem or not, it seems clear that if it is not increased it will be difficult to attract candidates to bank director positions given the additional duties expected of them in the future.

While representing a large fraction of banking industry assets, the sample in this paper is relatively small compared with the number of institutions in the financial sector. Furthermore, many more factors affect governance characteristics than the ones I considered here. Thus, much more research is needed to understand the extent to which governance contributed to the financial crisis. Nevertheless, the simple analysis in this paper is still suggestive that board-level governance of publicly traded financial firms may have played a part in the crisis. However, it also suggests that the recent governance reform movement embodied in SOX and the NYSE and Nasdaq listing standards may be as much to blame. SOX and the listing standards place a lot of emphasis on director independence. However, it is not clear if director independence is always beneficial because independent directors lack information (see also the arguments in Adams and Ferreira 2007). The problem may be exacerbated for financial firms because of the complex nature of their businesses. This is a point the OECD Steering Committee on Corporate Governance (Kirkpatrick 2009) also makes. It argues that independence at financial firms may have been overemphasized at the expense of qualifications. Guerrera and Larsen (2008) also discuss the fact that SOX made it more difficult for financial firms to hire suitable directors with financial expertise because of the perception of conflicts of interests.

Some tentative policy implications that can be drawn are as follows: By placing too much emphasis on independence, SOX and recent listing standards may have worsened board governance at publicly traded financial firms. Not only are financial firms complex, but the financial industry is complex because of the existence of different regulators. Until the governance of financial firms is better understood, it may be better not to impose restrictions on the governance of financial firms. Further regulating governance may have unintended negative side effects, as SOX may have had. These arguments suggest that, by emphasizing independence, Section 952 of the Dodd–Frank Act mandating fully independent compensation committees seems unlikely to prevent governance failures and may even contribute to them. Moreover, compliance with such regulations is costly. For example, Leuz et al. (2008) find that compliance with SOX was a significant factor in companies' decisions to delist in 2002–2004. These costs may be particularly large for small firms. Because the majority of firms in the banking sector are relatively small, at this point in time it is not clear how imposing additional restrictions on their governance can help prevent future crises.

It seems clear that it may be beneficial for financial firms to try to increase the financial sophistication of their boards, either through hiring new directors or through additional education. Because financial firms, such as Citigroup, already changed their boards, it is not clear that increasing the financial sophistication of the board requires regulation, for example, by allowing shareholder access to the proxy. Regardless of their expertise, it is likely that directors of financial firms will not need any regulatory prodding to ask more tough questions of management in the future.

One problem that remains is the issue of director compensation. To ensure that financial firms retain good directors and attract good candidates, directors of financial firms should be adequately compensated for the difficulties of their duties and the additional costs they bear in undertaking any additional training. However, it may be difficult to institute pay raises for directors of financial firms given the outrage over executive compensation following the crisis. It remains to be seen whether the provisions for greater oversight of executive compensation provided for in the Dodd–Frank Act will have a positive influence on director compensation.

  1. 1

    La Porta et al. (1998) create this index for 49 countries. The other countries with scores as high as the United States are Canada, Chile, Hong Kong, India, Pakistan, South Africa, and the UK.

  2. 2

    Regardless of exchange listing, all public companies are supposed to abide by SOX which requires that boards are responsible for internal control, audit committees consist entirely of independent directors, audit committees have at least one financial expert, management certifies financial statements, and board members face large penalties for corporate accounting fraud.

  3. 3

    But, it also identifies governance in large, complex, nonfinancial firms as a problem.

  4. 4

    See http://www.oecd.org/document/48/0,3343,en_2649_34813_42192368_1_1_1_37439,00.html.

  5. 5

    See http://www.hm-treasury.gov.uk/press_10_09.htm.

  6. 6

    Laeven and Levine (2009) provide insights into the governance role of bank ownership structure across various countries.

  7. 7

    Section 312 of the Dodd–Frank Act mandates the merger of the OTS into other regulatory bodies. The OTS will cease to exist on October 19, 2011.

  8. 8

    I use these data only after extensive cleaning.

  9. 9

    Some results in a previous version of this paper looked different because I did not cluster the standard errors at the firm level.

  10. 10

    http://www.ffiec.gov/nicpubweb/nicweb/NicHome.aspx.

  11. 11

    It is possible that regulators do not look favorably on pay raises for directors.

Ancillary