Endogenously structured boards of directors in banks

Electronic copy available at: http://ssrn.com/abstract=1251262

Endogenously Structured Boards of Directors in Banks

Shams Pathana* and Michael Skullyb

a UQ Business School, The University of Queensland, Brisbane, Queensland 4072, Australia.

b Department of Accounting and Finance, Monash University, Melbourne, Victoria 3145, Australia

Abstract

This paper examines the trends and endogenous determinants of boards of directors (board size,

composition, and CEO duality) for a sample of 212 US bank holding companies, from 1997 to

2004. Overall, the results show that the costs and benefits of boards’ monitoring and advising

roles could explain bank board structures with caveats. For example, due to the regulatory nature

and comparatively intensive scrutiny of bank officers and directors, it is argued that bank

managers have less control over the directors’ selection processes. Thus, bank board

independence should not be the outcome of negotiation with CEOs. Consistent with this view,

bank CEOs are found not to affect bank board independence. The trend analysis also provides

some important results. In contrast to non-bank evidence, for instance, board size was

discovered to decrease over the sample period for large and medium-sized banks, while board

size remained relatively stable for small banks. These results are robust with respect to different

estimation specifications. Furthermore, the study’s findings have important policy implications

for bank regulators and investors.

JEL classification: G21, G28, G30, G32, G34, L22, K22

Key words: Board of directors; Independent directors; CEO; CEO duality; Endogenous; Bank

Holding Companies

This paper is prepared from the first empirical chapter of the first author’s doctoral thesis at Monash University. The author wish to thank Barry Williams, Richard Heaney, Renée Adams, an anonymous referee, Ike Mathur (the Editor), Mark Flannery, Maureen O’Hara, Benjamin Hermalin, Robert Faff, John Kose, Mark Harris, Mohamad Ariff, J. Wickramanayake, Mamiza Haq, Peter Verhoeven, Janice How, Robert Bianchi, John Nowland, John Chen, Mohammad Hayat, Colleen Puttee, John Zhang, Alexander Akimov, George Tanewski, Petko Kalev, Hardjo Koerniadi, Mohamad Belkhir, and seminar/conference participants at Bond University, Deakin University, Queensland University of Technology, University of Western Sydney, Griffith University, 21st Australian Finance and Banking Conference, 2009 Asian Finance Association Conference, and 2009 AFAANZ Conference for their helpful comments. The author acknowledges the AFAANZ 2009 travel grant to support this research. The usual caveats apply.

* Corresponding Author: Email: [email protected]; Tel: 61 7 3346 8075

mailto:[email protected]


1. Introduction

A wide range of accounting, finance and management literature has determined that a

certain type of board structure is preferred to monitor managers. For instance, a small number of

board directors and more independent directors are considered to be important elements of an

effective board (e.g., Yermack 1996; Fama and Jensen 1983). This issue was further emphasized

by the introduction of the Sarbanes-Oxley Act of 2002 and the associated listing rules by NYSE,

NASDAQ, and AMEX as they require a majority of independent board directors and a

completely independent audit committee. Hence, these developments are in favor of a uniform

board structure, irrespective of the industry in question. However, if we believe Alchian’s (1950)

economic theory of ‘Darwinism,’ it is important to understand why some firms still maintain

large boards, while others have majorities of non-independent or executive directors. To answer

this question, several studies attempt to explain this observation by relating the costs and

benefits associated with boards’ monitoring and advising functions (Hermalin and Weisbach

1998; Raheja 2005; Adams and Ferreira 2007; and Harris and Raviv 2008). Based on these

theoretical works, among others, Bakers and Gompers (2003), Boone, Field, Karpoff, and Raheja

(2007), Coles, Daniel, and Naveen (2008), Linck, Netter, and Yang (2008), and Lehn, Patro, and

Zhao (2009) empirically find evidence in support of the endogenous formation of boards of

non-financial firms.

While the same theoretical underpinnings relating to board structure are valid to both

banks and non-bank firms, the existing empirical studies exclude banks from their sample, and

several factors (such as regulation, high leverage) could limit generalizing non-financial board

structure findings to banks. This study aims to fill this knowledge gap by investigating whether

the costs and benefits of the boards’ monitoring and advising functions could also explain board

structure (board size, composition, and CEO duality1) in a regulated industry, like banks.

The global financial crisis also highlights the importance of improving understanding of

bank governance. Indeed, the study on bank board structure deserves special attention for

several reasons. Perhaps the bank board of directors is even more important as a governance

mechanism than its non-bank counterparts because banks have become larger, complex and

more diversified, following the deregulation with the Riegle-Neal Interstate Banking and

Branching Efficiency Act of 1994, as well as the Gramm-Leach-Bliley Act of 1999. In addition,

the presence of regulation could have different implications for bank board structure

determinants. For example, as discussed later in Section 2, bank directors and managers are

1 ‘CEO duality’ refers to a situation in which the CEO is also the board chair.

1


subject to stringent regulatory scrutiny, compared to non-bank board directors. This regulation is

commonly justified for three reasons. First, there are costly consequences in the case of bank

failure (Flannery 1998). Second, bank shareholders have distorted objective of excessive risk-

taking in the presence of deposit insurance (Galai and Masulis 1976). Finally, bank debtors do

not have the incentive to monitor bank managers due to high information asymmetry

(Demirgüç-Kunt and Detragiache 2002). This constant regulatory monitoring could limit bank

managers’ self-serving behavior (such as perks). Hence, bank managers, including CEOs, cannot

influence the director selection process. As a result, in contrast to non-bank studies, CEO power

(i.e., CEO’s ability to influence board decisions) should not be an important determinant of bank

board independence. Thus, it is important to examine, even in the presence of such regulation,

whether bank board structure can still be explained by the costs and benefits associated with

boards’ monitoring and advising functions, given bank characteristics and other governance

mechanisms. It is also important to determine whether board structure has changed significantly

for regulated banks due to the enactment of the Sarbanes-Oxley Act (SOX) of 2002 and

associated listing rules changes. The study of the banking industry also provides a unique setting

in which to enhance our understanding of board structure determinants.

Using a sample of 212 US bank holding companies monitored between 1997 and 2004,

this study finds some evidence in favor of endogenously chosen boards of directors. This

supports the argument that banks structure their boards consistently with the costs and benefits

associated with boards’ monitoring and advising functions. More specifically, the results show

that: (i) larger and more diversified banks have larger and more independent boards, and also

combine both CEO and board chair titles; (ii) bank board independence is not the outcome of

negotiations with the CEO; (iii) banks in which managers’ opportunities to consume private

benefits are high have larger boards, while banks in which the cost of monitoring managers is

low have more independent boards; (iv) banks in which managers have substantial influence and

the constraints on managerial influence are weak combine both CEO and board chair roles; and

(v) banks in which insiders’ shareholding is high and the outsiders’ shareholding is low have

smaller boards.

The trends in bank board structure over the sample period also provide some significant

insights. For example, bank board size declines over the sample period, particularly for large and

medium size banks, which is in contrast with non-bank firm evidence. However, the percentage

of independent directors increases substantially, especially during the post-SOX period.

2

This study contributes to the existing literature on board structure determinants in

several important ways. This is the first study to demonstrate that even in a regulated industry

like banking, the costs and benefits of monitoring and advising functions of boards could explain

their structure. This paper complements and extends Adams and Mehran’s (2009) study, which

investigates bank board governance for a sample of 35 BHCs from 1959 to 1999. They illustrate

that bank board size relates to M&A activity and organizational structure. However, they have

not shown the determinants of other important board features, such as board composition and

leadership structure. Likewise, they have not exclusively examined the determinants of bank

board structure (such as negotiations with the CEO, ownership incentive structure), in view of

existing non-bank evidence by Lehn et al. (2009), Linck et al. (2008), and Boone et al. (2007),

among others. This study also broadens our knowledge by showing that bank CEOs do not

influence the board selection process due to fear of regulatory action. This result challenges the

existing non-bank evidence and thus has important policy implications, while designing

appropriate governance system for banks. In terms of methodology, a broad set of diagnostic

and statistical consistency tests were conducted to confirm the robustness of the results,

including several approaches that account for unobserved heterogeneity and simultaneity. For

example, a system generalized method of moments (GMM) estimation technique was used to

directly control for any ‘dynamic endogeneity' problem. Finally, this is perhaps the first study to

provide some evidence that bank board structure has significantly changed, following SOX and

the associated changes mandated by the stock exchanges. Such findings are vital to evaluating the

possible impact of SOX on regulated banks’ board structure.

The rest of the paper is structured as follows. Section 2 further drives the study of bank

board structure determinants by discussing the regulatory oversight of boards of directors in

banks. Section 3 reviews the literature on board structure determinants and formulates the

relevant hypotheses for banks. Then, Section 4 describes the data and methodology. Section 5

provides the empirical results, while Section 6 demonstrates the robustness of the results, using

different estimation techniques. Section 7 reports some results with regard to the impact of SOX

and associated listing rules’ changes on bank board structure determinants. Finally, Section 8

concludes the paper.

3

2. Regulatory oversight of boards of directors in banks

Banks’ boards of directors historically have not been legally bound to solely serve the

shareholders, as is typically the case for non-bank firms. The ‘fiduciary’2 responsibility (i.e., duty

of loyalty and care) of the bank directors and managers extends beyond shareholders to

depositors and bank regulators (for more details, see Fisher 1992; Macey and O’Hara 2003;

Fanto 2006). Bank regulators set detailed standards of conduct for directors and managers and

monitor individual conformity with these standards to ensure ‘safe and sound’ bank system. The

regulators have considerable disciplinary powers available, if they discover bank directors and

managers in any violation of the standards. The disciplinary actions include suspension and

removal from the bank, and even a life-long ban from the industry; regulators can also refer the

matter to federal prosecutors. With the passing of the Financial Institutions Reforms, Recovery

and Enforcement Act (FIRREA) of 1989, and the FDIC Improvement Act of 1991, Congress

further empowered bank regulators in taking ‘prompt corrective actions’ against bank directors

and officers for their decisive roles (see Shepherd 1992 for details). For example, Section 1821(k)

of FIRREA 1989 stated that directors and officers of insured institutions would be held

personally liable for any misconduct of bank business (Shepherd 1992, p. 1122).

The available data indicate that bank regulators frequently use these disciplinary powers

against bank directors and managers. For example, in 2005, of the 32 consensual removal orders

by the Office of the Comptroller of the Currency (OCC), 12 involved senior bank officers,

including CEO and directors.3 Thus, bank directors and managers have a legal duty to recognize

their obligation to the debt claimants, especially when a bank is in a weakened financial

condition. Fanto (2006, p.6) uses the metaphor that “bank regulators stay with bank directors

and officers from the cradle to the grave” to illustrate how intensively bank regulators monitor

and discipline bank management.

It should also be noted that certain regulations at the bank level, as opposed to the

BHC’s level, can constrain bank board size and composition. For example, the board of a

national bank (regulated and supervised by the OCC) must consist of five to twenty-five

directors. However, the comptroller can hold the national bank exempt from such a limit. The

majority of directors of national banks must also be selected from a certain proximity to the

2 Macey and Miller (1993, p. 401, 407) define fiduciary duties as “. . . the mechanism invented by the legal system

for filling in the unspecified terms of shareholders’ contingent [contracts].” In addition, see Macey and O’Hara (2003, pp. 93-95) for a good discussions of bank directors’ fiduciary duties, or ‘duty of care’ and ‘duty of loyalty’ to shareholders, depositors and regulators.

3 Source: The OCC Web site < http://apps.occ.gov/EnforcementActions/> (viewed on August 27, 2009) for a search engine for enforcement actions.

4

http://apps.occ.gov/EnforcementActions/

bank’s head office, unless the residency requirement is waived by the comptroller. The bank

directors are required to have invested funds in their banks. Similarly, boards of state member

banks are subject to specific state government directives. For example, New York State requires

its member banks to maintain boards of seven to thirty directors (when their capital stock,

surplus, and divided profits are in excess of $50 million), while two-thirds of these directors

should be non-executives, i.e., outsiders.4 During this study’s sample period, the New York State

member banks were required to hold a minimum of ten meetings per year (two conference call

meetings were allowed). State regulations on the number of meetings could influence the bank’s

choice of directors, as those living within proximity might be more likely to be selected.

Although boards of national and state banks are required to meet such regulatory

standards, BHCs’ boards–the focus of this study–are exempt from such state requirements. The

Federal Reserve System acts as the ‘umbrella supervisor’ for BHCs and do not impose any such

specific requirements on BHCs’ boards. Therefore, the regulatory environment alone cannot

fully explain BHC board structure.

3. Literature review and hypotheses’ development

Prior studies argue that optimal board structure is based on the costs and benefits of the

board monitoring and advising roles, along with other firm and governance characteristics (Linck

et al. 2008, p. 311). The two most important roles of a board of directors are monitoring and

advising (e.g., Jensen 1993; Raheja 2005; Adams and Ferreira 2007; Linck et al. 2008). As a

monitor of managers, the board supervises the management so as to refrain them from any self-

serving behaviors, such as shirking and perquisites. In its advising role, the board provides

opinions and directions to managers for key strategic business decisions. Typically, in explaining

boards, previous studies model two specific elements of the board: board size and board

composition (i.e., independent directors) as points of reference. Indeed, the theoretical

arguments on board structure determinants can be extended to explain other board structure

variables. Accordingly, the following sections, 3.1 to 3.4, elaborate on the ‘scope of operations’,

board monitoring requirements, CEOs ‘negotiations’ power and their succession process, and

ownership incentives structure as possible determinants of board structure (board size,

independence, and CEO duality).5

4 Source: www.banking.state.ny.us 5 It is important to note that the terms used to explain the different determinants of board structure are from the

relevant literature, to avoid any confusion.

5

3.1 Scope of operations

The term ‘scope of operations’ refers to the nature, diversity and complexity of the

business production process (Boone et al. 2007; and Linck et al. 2008). In comparison to small

firms, large and diversified firms require additional board members to support their complex and

diversified activities (Agrawal and Knoeber 1996; Bhagat and Black 2002), and also to monitor

management performance (Coles et al. 2008; Lehn et al. 2009). Large firms could also require

more directors to serve on their board’s sub-committees, handling the nomination of board

members, compensation, and auditing (Boone et al. 2007). Similarly, the information

requirements of larger and more complex firms generally result in the need for larger boards. In

this regard, prior studies have established a positive relationship between board size and the

firm’s ‘scope of operations’ (e.g., Boone et al. 2007; Coles et al. 2008; Linck et al. 2008; Lehn et

al. 2009).

In addition to board size, the firm’s ‘scope of operations’ could also affect the board

composition, i.e., board independence (Boone et al. 2007; Coles et al. 2008; Linck et al. 2008;

Lehn et al. 2009). Since outside independent directors are better monitors, large and complex

firms could require more of them so as to reduce the augmented agency problems of being large

(Lehn et al. 2009). In addition, Fama and Jensen (1983) and Linck et al. (2008) consider outside

directors to be of high importance to large and complex firms, since they bring valuable expertise

and potential networks that could be beneficial to the firm. They further contend that even

though ‘monitoring costs’6 increase with a firm’s ‘scope of operations’, the benefits from

effective monitoring offset the costs ‘on balance’. These arguments are consistent with those of

Boone et al. (2007), Coles et al. (2008) and Linck et al. (2008), all of whom hold that board

independence is positively related to the scope of operations. To capture the different aspects of

the so-called scope of operations, prior studies have used multiple proxies for it, such as firm

size (i.e., total assets), age, leverage and the number of business segments involved (Boone et al.

2007; Linck et al. 2008).

Thus, for large and diversified banks, additional board members and possibly more

independent directors are required to monitor management (Boone et al. 2007, p.70) and to

advise on new product markets, technology, regulations, M&A and so forth (Lehn et al. 2009,

p.4). Based on these arguments, the first hypothesis H1 related to the ‘scope of operations’ is as

follows:

6 The term ‘monitoring costs’ is used to describe the costs related to information acquisition and processing in

transforming directors’ expertise to the specific firms for which they serve as directors (Linck et al. 2008, p.311).

6

Hypothesis H1: Bank board size and the percentage of independent directors are

positively related to the bank’s scope of operations.

3.2 Board monitoring requirements

The term ‘board monitoring requirements’ is used to express that board structure is also

affected by the net benefits of monitoring managers’ ‘private benefits,’7 as well as the

‘monitoring costs’ to directors (Raheja 2005; Adams and Ferreira 2007; Harris and Raviv 2008).

With regard to ‘private benefits’, the benefit gained from monitoring managers increases if

managers have opportunity to extract more ‘private benefits’ from their firms (Boone et al.

2007). Generally, for managers, such opportunities arise when firms have free cash flows (Jensen

1993), and managers are immune to any shareholders’ activism, i.e., M&A activities (Boone et al.

2007). For instance, if boards are ‘staggered,’8 then shareholders or potential acquirer could not

be able to discipline managers, as it will be difficult, if not impossible, to remove all the directors

at the same time. In the presence of such opportunity for greater ‘private benefits’ to insiders,

boards will hire more independent directors and so increase in overall size (Boone et al. 2007,

p.71).

With regard to ‘monitoring costs,’ these are greater for firms with high information

asymmetry (Fama and Jensen 1983; Jensen 1993; Gillan, Hartzell and Starks 2006). Prior studies

suggest that firms with greater ‘monitoring costs’ should rely less on outside directors (Fama and

Jensen 1983; Lehn et al. 2009). It is costly to transfer firm-specific information to outsiders

because they are less informed about the firm’s projects (Linck et al. 2008, p. 311). In addition,

large boards could have less motivation (to incur additional costs or efforts) to acquire

information due to ‘free-riding’ problems, as well as higher coordination and direct costs, such as

remuneration (Linck et al. 2008). In contrast, inside directors have access to firm-specific

information as part of their day-to-day activities. Thus, theoretical models of Raheja (2005),

Adams and Ferreira (2007) and Harris and Raviv (2008) on board structure predict that the

number of outsiders decreases with ‘monitoring costs.’ Therefore, firms with high information

asymmetry could benefit from smaller board size and a greater representation of inside directors

because of high monitoring costs. Generally, information asymmetry is high for firms with high

stock return volatility (Fama and Jensen 1983; Gillan et al. 2006), high growth potential (Jensen

7 The term ‘private benefits’ is defined as the insiders’ efforts-aversion, perks from inferior projects and

opposition to acts against the CEO (Raheja 2005, p.298). 8 A board is staggered when its directors are divided into classes, generally three classes, with only one class of

directors standing for re-election each year. Thus, shareholders cannot replace the majority of directors in any given year, even though there could be widespread shareholder support for such a change.

7

1993; Gillan et al. 2006; Boone et al. 2007; Linck et al. 2008), and high R&D expenditures

(Boone et al. 2007; Coles et al. 2008).

Thus, an optimal board will have more outside independent directors and be larger in

overall size, when management ‘private benefits’ are high and the cost of monitoring is low.

Accordingly, Boone et al. (2007) find a statistically significant positive (negative) relation between

monitoring ‘private benefits’ (‘monitoring costs’) and board size, but not the same for board

independence. Meanwhile, Linck et al. (2008) and Lehn et al. (2009) support the negative impact

of ‘monitoring costs’ on both board size and board independence. Linck et al. (2008) do not

explore the effect of monitoring ‘private benefits’ on board size, but instead find a statistically

significant positive relation between monitoring ‘private benefits’ and board independence. Thus,

based on this discussion, the second hypothesis related to bank board monitoring requirements

is as follows:

Hypothesis H2: Bank board size and the percentage of independent directors are

positively related to management private benefits and negatively related to directors’

monitoring costs.

3.3 CEOs negotiations power and their succession process

The ‘negotiation’ theory illustrates that board structure, particularly board independence,

is the outcome of negotiations between the board and CEOs (Hermalin and Weisbach 1998).

That is, board independence decreases with CEO’s (negotiation) power and increases with

constraints on such CEO power (Hermalin and Weisbach 1998; Baker and Gompers 2003;

Boone et al. 2007; Linck et al. 2008). CEO (negotiation) power generally derives from the CEO’s

perceived ability (relative to a replacement) to influence board decisions, as can be proxied by

firm performance, CEO tenure or CEO ownership. Likewise, CEO’s (negotiation) power can be

constrained by the presence of shareholdings of non-affiliated block-holders or outside directors

on the board (Boone et al. 2007).

Board independence could also be influenced by the CEO succession process (Hermalin

and Weisbach 1998; Linck et al. 2008). The two most common types of CEO-succession

processes are ‘horse races’ and ‘passing the baton’ (Mace 1971; Vancil 1987; and Brickley, Coles

and Jarrell 1997). Using the ‘horse races’ argument, the firm conducts a tournament among

eligible candidates for the CEO position (Brickley et al. 1997). Under the ‘passing the baton’

argument, the board chooses a designated successor for the CEO in advance (Mace 1971; and

Vancil 1987). This succession process indicates that board independence decreases as the CEO

8

approaches retirement (Mace 1971; and Vancil 1987), as is proxied by CEO age (Linck et al.

2008).

However, for banks–due to statutory and regulatory considerations, as mentioned earlier

in Section 2–bank management, including CEOs, neither can influence the directors’ selection

processes nor they have any succession planning in fear of severe penalties for any misconduct.

For example, on November 24, 2003, Richard M. Thomas, former CEO and President of First

National Bank, was disciplined by OCC with an industry ban, $50,000 in restitution, transfer of

shares, and a $10,00 civil money penalty for his misconduct (OCC No. 2003-106). Likewise, on

December 22, 2003, Eduardo Masferrer, former CEO and board chair of Hamilton Bank, was

disciplined with an industry ban, $960,000 in restitution, and a $40,000 civil penalty (OCC No

2003-150). Therefore, in contrast to non-bank firm evidence, this study reasonably predicts that

bank board independence is neither negatively related to CEO negotiation power and closeness

of CEO retirement, nor positively related to constraints on such CEO power. Thus, the third

hypothesis related to the CEO negotiation theory and succession process is stated in an

alternative format as:

Hypothesis H3: The percentage of independent directors is negatively related to CEO

power and closeness of CEO retirement, and positively related to constraints on such

CEO power.

3.4 Ownership incentives structure

The board could also reflect the firm’s ‘ownership incentives structure’ (Rediker and Seth

1995; Raheja 2005; Linck et al. 2008). This ‘ownership incentives structure’ explains that

variations in the firm’s ownership structure can also be important in aligning both managers’ and

shareholders’ interests (e.g., Morck, Shleifer and Vishney 1989; McConnell and Servaes 1990).

These incentives may substitute or complement the board as internal governance mechanisms

and hence can be considered as a relevant determinant of board structure. Raheja (2005)

contends that when both shareholder and management incentives are aligned, boards will be

smaller. This is because there is a low demand for outside monitors, when insiders are less likely

to select inferior projects. Linck et al. (2008) use this notion to support the negative relationship

between the CEO shareholdings (as proxy for insider incentives) and board size.

Raheja (2005) further argues that greater outside directors’ shareholdings raises their

incentives to monitor management, as well as to verify the projects more vigorously because of

their increased shares of the firm’s profit. These shareholdings reduce project verification costs,

and potentially communication and coordination costs as well. Raheja’s (2005, Proposition 6, p.

9

296) model, therefore, specifies that an optimal board is larger and has a majority of independent

directors when verification costs are low. However, Linck et al. (2008) find that outside directors’

shareholdings decrease both board size and board independence. They reason that fewer outside

directors may be required when each director has stronger incentives to monitor management.

Likewise, for banks, Rediker and Seth (1995) find that the proportion of outside directors is

negatively related to both the non-affiliated block-holders’ shareholdings and the managerial

shareholdings. However, consistent with Raheja’s (2005) proposition for banks, Whidbee (1997)

shows that the proportion of outside directors is negatively related to CEO ownership, while

positively related to non-affiliated block ownership. Based on this discussion, the fourth

hypothesis with regard to ownership incentive alignment is as follows:

Hypothesis H4: Bank board size is negatively related to insiders’ incentives alignment and

positively related to outsiders’ incentives alignment.

3.5 CEO duality determinants

The existing theoretical studies do not explicitly address the determinants of CEO

duality. However, their implications–along with existing empirical findings–can nevertheless help

in formulating some specific expectations regarding CEO duality (e.g., Hermalin and Weisbach

1998; Linck et al. 2008). As an insider, the CEO possesses firm-specific knowledge that is

important for large, complex and diversified banks. Likewise, banks with high monitoring costs,

i.e., with high information asymmetry, could benefit from CEO duality. It could also be argued

that in the presence of opportunities for insiders to extract private benefits, the CEO and board

chair roles should be separated to achieve a balance between board independence and such

opportunities. Based on these arguments, the following two hypotheses, H5A and H5B, for bank

CEO duality determinants relate to ‘scope of operation and board monitoring requirements

respectively:

Hypothesis H5A: CEO duality is positively related to the bank’s scope of operations.

Hypothesis H5B: CEO duality is negatively related to management private benefits and

positively related to directors’ monitoring costs.

Consistent with Hermalin and Weisbach’s (1998) ‘negotiation’ theory, as explained in

Section 3.3, it can be argued that a CEO with greater power will also chair the board, so as to

gain more power. Similarly, constraints on CEO power will favor separating the CEO and board

chair positions. Compatible with Brickley et al. (1997) succession planning theory, CEOs are

honored with board chair title as they approach retirement. In support of these notions, Linck et

al. (2008) find that the probability of CEO duality increases with CEO power and closeness of

10

CEO retirement, and decreases with constraints on CEO power. Thus, hypothesis H5C for bank

CEO duality determinants related to the CEO negotiation theory and succession process can be

specified as follows:

Hypothesis H5C: CEO duality is positively related to CEO power and closeness of CEO

retirement, and negatively related to constraints on such CEO power.

4. Data and empirical method

4.1 Data and sample procedure

The initial sample consists of the largest BHCs headquartered in the United States with

standard industrial classification (SIC) of 6021 and 6022 for respective national and state

commercial banks, over the period from 1997 to 2004. The data are sourced from DEF 14A

proxy statements, BANKSCOPE, FR Y-9C, DATASTREAM, and SDC Platinum.

Detailed information on bank board structures is hand collected from DEF 14A proxy

statements of annual meetings found in the SEC’s EDGAR filings. Following Adams and

Mehran (2009), the governance data are measured on the date of the proxy statement, i.e., at the

beginning of the respective fiscal year. The data collection procedure is then adjusted to account

for when the proxies disclose some governance information for the previous fiscal year (e.g., the

percentage of CEO shareholding) and others for the following fiscal year (e.g., the number of

directors). The financial information on BHCs are mostly obtained from the BANKSCOPE

database and complemented by the fourth quarter Consolidated Financial Statements for BHCs,

i.e., Form FR Y-9C, from the Federal Reserve Board. The market information on BHCs is

collected from the DATASTREAM database. Similarly, the US three-month Treasury bill rate in

the two-index market model for bank risk computations is obtained from the Federal Reserve

Bank of St. Louis. The information on M&A activities of the sample BHCs over the sample

period are obtained from Thomson Financial’s SDC Platinum database. The initial sample begins

with the three hundred largest BHCs, as ranked by the 2004 year-end book values of total assets.

The final sample, an intersection of the data on BHCs with SIC 6021 and 6022 in DEF 14A

proxy statements, BANKSCOPE, DATASTREAM, and with a minimum two consecutive years’

data between 1997 and 2004, consists of 1,534 observations on 212 BHCs.

4.2 Measures of variables

The three left-hand side variables, i.e., the board structure variables, explained in this

paper are board size (BS), independent director (INDIR), and CEO duality (DUAL). BS is the

total number of directors serving on the bank board at the end of each fiscal year. INDIR is the

number of independent directors, as a percentage of the total number of board directors.

11

Following prior studies (e.g., Brickley, Coles and Terry 1994; Hermalin and Weisbach 2003;

Adams and Mehran 2009), this study defines independent directors as those whose only business

relationship with the bank is their directorship. When evaluating director’s independence,

following Adams and Mehran (2009), the borrowing and depositing by directors with their BHCs

or their subsidiaries are also considered.9 DUAL is a dummy variable that equals one if the CEO

also chairs the board, but is otherwise zero.

The measurements of the four explanatory variables (i.e., determinants)–related to the

testing hypotheses about scope of operations, monitoring requirements, negotiations with CEOs

and ownership incentives structures–are as follows. Bank ‘scope of operations’ is proxied by

three variables: bank size, bank age, and revenue diversification index. Bank size (TA) is the

bank’s total assets, and bank age (AGE) is the number of years since a bank was first listed on

DATASTREAM. Following Stiroh and Rumble (2006, p. 127), bank revenue diversification

index (DIVER) is computed as one minus the sum of the squared fraction of operating income

from interest and the squared fraction of net operating income from non-interest sources. Prior

studies, including Lehn et al. (2009), Linck et al. (2008), and Boone et al. (2007), used the number

of business segments as a measure of diversification for non-bank firms. However, for banks,

Stiroh and Rumble’s (2006) revenue diversification approach seemed more appropriate because

it captures the complexity and the level of diversification of banks through their income sources.

DIVER measures the degree of diversification in a BHC’s net operating revenue. Prior studies,

including Lehn et al. (2009), Linck et al. (2008), and Boone et al. (2007), used the number of

business segments as a measure of diversification for non-bank firms. However, for banks,

Stiroh and Rumble’s (2006) revenue diversification approach seemed more appropriate because

it captures the complexity and the level of diversification of banks through their income sources.

A shareholder’s restrictive rights index (EINDEX), an approximation of Bebchuk,

Cohen and Ferrell’s (2009) entrenchment index,10 is used as a proxy for managers’ opportunities

for private benefits. EINDEX is computed as the sum of two dummy variables: staggered board

(STAGG) and poison pill (POISON). The dummy variable, STAGG, equals one if the board is

staggered; otherwise, it is zero. The dummy variable, POISON, equals one if the bank board has

the provision for poison pill, but is otherwise zero. The two proxies for directors’ monitoring

9 Any loans should comply with the applicable law, including Regulation “O” of Board of Governors of Federal Reserve and Section 13(k) of the Securities Exchange Act of 1934.

10 Bebchuk et al. (2009) put forth the entrenchment index as the composite of six dummy variables: staggered boards, limits to shareholder by-law amendments, supermajority requirements for mergers, supermajority requirements for charter amendments, poison pills and golden parachutes. Definitions of these variables are in the study by Bebchuk et al. (2009, p. 824). In an unreported correlation between a Bebchuk entrenchment index and EINDEX is 0.67 (p-value < 0.01) for a sample of 49 banks.

12

costs are bank charter value (CV) and bank risk (RISK). CV is calculated following Keeley’s Q

(1990) as the ratio of the market value of total assets to the book value of total assets, while the

market value of total assets is the sum of market value of equity plus book value of liabilities. All

these values are measured at the end of fiscal year. It is important to note that CV itself is

considered to be an important controlling device for banks, as it is affected by regulation and

hence alters the need for monitoring by independent directors (Booth, Cornett and Tehranian

2002). High charter value could reduce moral hazard problems in banks because shareholders in

banks with high charter values have more to lose if a risky project becomes insolvent. Thus,

charter value reduces the need for independent directors as a monitoring device (Marcus 1984).

RISK is calculated as the standard deviation of the bank’s daily stock return for each year.

CEO power is measured by three proxies: the prior period return on average (ROA)

which is the net income after tax as a percentage of average total assets; CEO tenure (CEOT),

which is the number of years spent as the bank CEO; and the percentage of the bank’s total

outstanding shares owned by the bank CEO (CEOWN). CEO succession planning is proxied by

the CEO’s age in years (CEOAGE). Likewise, following Boone et al. (2007) and Linck et al.

(2008), constraints on CEO power are measured by two variables: outside ownership

(OUTSIDEOWN) and non-affiliated block ownership (NBLOCKOWN). OUTSIDE is

calculated as the number of shares held by the bank officers and directors, excluding the CEO,

as reported in DEF 14A proxy statements as a percentage of the total number of outstanding

shares.11 Finally, NBLOCKOWN is calculated as the number of shares owned by non-affiliated

block holders who own 5% or more shares as a percentage of total number of outstanding

shares.12 Non-affiliated block holders exclude inside directors and any trust company holding

stocks on behalf of the bank’s employee stock ownership plan because the bank managers may

have influence over those trust companies through the contractual nature of their relationships.

Several additional variables are included as control variables to reduce biases in the

coefficient estimates due to omitted variables. These additional bank-specific control variables

include: bank capital ratio, dummies for merger, post-SOX period, and lag of BS and INDIR.

The bank capital ratio (CAPITAL) is measured as bank total equity as a percentage of the bank’s

11 Boone et al. (2007) use the percentage of total outstanding shares owned by outside directors, rather than by

officers and directors, as a proxy for outsiders’ incentive alignment. Board ownership can be a reasonable proxy for outside director ownership because the correlation between these two variables is 0.74 and statistically significant at 1% level for a random sample of 40 banks.

12 SEC requires banks to disclose those shareholders owning at least five percent of the firm’s total outstanding shares in the DEF 14A proxy statement. The shareholdings of others with less than five percent are also sometimes reported. However, in the determinant analysis for BS and BM, NBLOCKOWN is omitted, as it proved to be highly correlated with OUTSIDEOWN (-.1015 with p-value < 0.01).

13

total assets. CAPITAL is expected to positively affect both BS and INDIR because a high capital

ratio means a lower level of debt. Debt, such as subordinated debt, is considered to be an

important market monitoring mechanism in disciplining bank managers (Flannery and Sorescu

1996). Hence, given the absence of such monitoring mechanisms, other internal governance

techniques, such as independent directors, may become more important. The dummy for M&A

(MERGER) equals one for a bank that completed any M&A activity during that year, but is

otherwise zero. MERGER is included to control for any prior period M&A activity (if any) by a

bank because any recent M&A activity could affect bank board structure (Boone et al. 2007;

Adams and Mehran 2009). For instance, the positive association between bank board size and

prior period M&A activity might reflect the addition of directors from an acquired bank’s board

into the merged/acquirer bank’s board. The dummy for the post-SOX period (DSOX) equals

one if the period is either 2003 or 2004; otherwise, it equals zero. It is included to control for the

post-SOX environment and also to capture the impact of SOX on the board variables. The

coefficient on DSOX also complements the univariate test results in Section 6, showing if bank

board variables changes were statistically significant in the post-SOX period. Finally, following

prior studies (e.g., Boone et al. 2007; Linck et al. 2008), lags of BS and INDIR are included in the

respective regression equations to capture the interaction between different board structure

variables, i.e., to reduce endogeneity of the variable of interest.

4.3 Empirical models and estimation methods

4.3.1 Empirical models

The following three regression equations, Equations (1) to (3), are specified to test

formally the determinants hypotheses, respectively, for BS, INDIR, and DUAL, given the

theoretical and empirical discussion in Section 3. The hypothesized signs of the equations are

shown beneath the respective variables in the following equations.

⎪⎪

⎩

⎪⎪

⎨

⎧

++

+++++

++++++

=

++−

+−++−

−−++++

tititi

titititi

titititititi

ti

DSOXMERGER

CAPITALINDIROUTSIDEOWNCEOWN

RISKCVEINDEXDIVERAGETA

BS

,,1,3

,21,1,2,1

,3,2,1,3,2,1

,

)()(

)()()()(

)()()()()ln()ln(

)ln(

εϕζ

ζζγγ

δδδβββα (1)

⎪⎪

⎩

⎪⎪

⎨

⎧

+++++

+++++

+++++++

=

++−+−++

+−−−−−

−−++++

titititititi

tititititi

titititititi

ti

DSOXMERGERCAPITALBSNBLOCKOWN

OUTSIDEOWNCEOAGECEOWNCEOTROA


INDIR

,,1,3,21,1,2

,1,4,3,21,1

,3,2,1,3,2,1

,

)()()ln()ln()(

)()ln()()()(

)()()()()ln()ln(

)(

εϕζζζλ

λφφφφ

δδδβββα (2)

14

⎪⎪

⎩

⎪⎪

⎨

⎧

+++++

+++++

+++++++

=

−−−+−−−

−+++−+

++−+++

titititititi

tititititi

titititititi

ti

DSOXMERGERCAPITALINDIRNBLOCKOWN

OUTSIDEOWNCEOAGECEOWNCEOTROA


DUAL

,,1,3,21,1,2

,1,4,3,21,1

,3,2,1,3,2,1

,

)()()ln()()(

)()ln()()()(

)()()()()ln()ln(

)(

εϕζζζλ

λφφφφ

δδδβββα (3)

where subscripts i denotes individual BHC (i = 1, 2, …, 212), t refers to time period (t =

1997, 1998, ….., 2004), and ln is the natural logarithm. β, δ, φ, λ, γ, ζ, φ, and Ψ are the parameters

to be estimated. ε is the idiosyncratic error term. The definitions of the variables and the relevant

hypothesized (or expected) signs in regression equations (1) through (3) are as already discussed

in Sub-section 4.2 and summarized in Table 1.

[INSERT TABLE 1 ABOUT HERE]

4.3.2 Estimation methods

Following prior studies, including Boone et al. (2007), Coles et al. (2008), and Linck et al.

(2008), the primary estimation method of regression equations (1) and (2) for board size (BS) and

independence (INDIR), respectively, is pooled ordinary least squares (OLS). Due to the binary

nature of the variable, regression equation (3) for CEO duality (DUAL) is estimated with the

maximum-likelihood LOGIT model. A priori, the variance-covariance matrix in the pooled-OLS

estimates will be adjusted with Huber (1964) or White’s (1980) estimators, which are robust with

respect to heteroskedasticity. Adopting Petersen (2009) procedure, observations are also

clustered by both panels (i.e., by banks) and time period to address random unobserved serial

and cross-sectional correlation respectively (if any) in residuals.

This study applies several measures to reduce endogeneity in the right-hand side

variables. For example, lagged values of both BS and INDIR are included as instrumental

variables in the respective board structure determinants regression equations. As an additional

robustness check, simultaneous equation model (three-stage least square (3SLS)) is also estimated

and reported in Section 6.1, in which structural models are specified as endogenizing board

structure variables. Particularly, Equations (1) to (3) for BS, INDIR, and DUAL respectively are

all estimated in a simultaneous system. System GMM estimates in Section 6.2 are also robust to

unobserved heterogeneity, simultaneity and dynamic endogeneity (if any).

4.4 Descriptive statistics and correlation analysis

The descriptive statistics for the various board structures, CEO characteristics,

ownership, and bank-characteristics variables are presented in Table 2. The board structure

variables in Panel A of Table 2 show that the mean (median) number of bank board directors,

BS, is 12.92 (12.00), with a minimum of 5 and a maximum of 31. The mean (median) percentage

15

of independent directors, INDIR, of 64.52 (66.67%) is similar to the 66.52% reported by

Cornett, McNutt and Tehranian (2009) for banks. Seventy-four percent of the sample banks

have staggered boards, and thirty-four percent have a poison pill provision. The mean value of

shareholders’ restrictive right index (EINDEX) is 1.08.


In Panel B of Table 2, the descriptive statistics of CEO characteristics indicate that 58%

of the sample banks combined both CEO and board chair titles (DUAL). The mean (median)

tenure of the CEO, CEOT, is 8.85 (7.00) years, which is consistent with the mean CEO tenure

of 7.44 in Houston and James’s (1995) sample of banks. The mean (median) age of the CEO,

CEOAGE, is 56.09 (56.00) and consistent with that reported by both Houston and James (1995)

and Cornett et al. (2009). The mean (median) CEOWN is 4.41% (1.30%). In Panel C of Table

3.5, within the two ownership variables, the mean (median) OUTSIDEOWN is 10.25% (7.24%),

which is comparable to the 9.63% reported by Anderson and Fraser (2000). The mean (median)

NBLOCKOWN is 3.67% (7.24%), which is lower than that of 6.64%, as reported by Anderson

and Fraser (2000). For brevity, the descriptive statistics of ownership and bank-specific variables

(in Panels C and D of Table 2) are not described further.


Table 3 presents the Pearson product-moment correlation coefficients among the

governance and bank-specific variables. It advances some initial guesses about the determinants

of board structure. For example, the highly positive correlation between bank board size (BS)

and bank size (TA) indicate that board size increases with bank size. It also indicates that the

multicollinearity could be a concern in the multivariate analysis. For instance, TA, DIVER and

AGE are highly positively correlated.

4.5 Trends in bank board structure

To place the study in comparison to non-bank studies, Figure 1 below shows the time

trends of bank board structure (i.e., board size, percentage of board independence, and CEO

duality) from 1997 to 2004. The three size groupings are developed by ranking the banks into

quartiles based on their total assets in each year. Then, the first quartile banks are grouped as

small, the second and third quartiles as medium, and the fourth quartile large.

[INSERT FIGURE 1 ABOUT HERE]

Panel A shows the trend in the mean bank board size. Consistent with the non-bank

evidence, larger banks have larger boards. Bank boards generally decrease over the sample

16

period, except for during 1997 and 1998. This decrease is most remarkable for large and medium

banks. Panel B reports the time trend in the mean percentage of independent directors. The

larger banks appear to have more independent directors than medium and small banks. The

percentage of independent directors remains relatively flat for both large and medium banks until

2001, and then increases slightly in 2004. However, small banks exhibit the most dramatic swing

in the percentage of independent directors over the sample period. For small banks, it declines

from 64.18% in 1997 to 55.66% in 2000, and then increases to 61.98% in 2004. Finally, Panel C

shows that as with board size and independence, bank size is an important determinant of CEO

duality, i.e., whether banks combine both CEO and board chair positions. Similar to non-bank

evidence, we can see that there is no strong time trend in terms of bank CEO duality.

5. Empirical results

Table 4 below reports the results of regression equations (1) to (3). The relevant

diagnostic tests in Panel C of Table 4 are based on pooled-OLS without any robust adjustment

for residuals. The average variance inflation factors (AVIF) across all the columns indicate that

the multicollinearity among the regressors should not be a concern in estimating the regression

equations.13 The White (1980) alternative test for heteroskedasticity (Π1) shows statistically

significant LM-statistics for each regression, which confirms the presence of heteroskedasticity

with normal OLS estimates. Likewise, the pooled-OLS estimates appear to suffer from first-

order serial correlation, as indicated by statistically significant F-statistics across all regressions

with Wooldridge’s (2006) test for first-order serial correlation (Π2). The presence of first-order

serial correlation in the panel data also indicates the presence of an ‘unobserved firm-fixed effect’

(Wooldridge 2002, p.176). These justify the pooled-OLS estimates of equations (1) and (2) and

LOGIT estimates of equation (3), with Huber (1964) or White (1980) heteroskedasticity robust

standard errors. The observations are also clustered by both banks and times to control for

unknown fixed- and time-effects in the estimates. In Panel B of Table 4, the regression equations

are well fitted with adjusted/pseudo R-squared of 27.55%, 21.25%, and 16.74%, respectively, for

BS, INDIR, and DUAL regressions, with statistically significant F-statistics.


With regard to hypothesis H1, as mentioned earlier in Section 4.2, three variables are used

to proxy for the bank’s ‘scope of operations’ TA, AGE, and DIVER. Even with the criticism of

‘attenuation bias’ due to the inclusion of multiple proxies for one underlying variable, the

13 According to Chatterjee, Hadi and Prince (2000), the guidelines for detecting multicollinearity are: (i) the largest

VIF is greater than 10, and (ii) the mean VIF is larger than 1.

17

coefficients on TA and DIVER remain statistically significant in the BS regression. The use of

multiple proxies for an underlying variable in one regression equation could bias the coefficient

toward zero, which is commonly known as ‘attenuation bias’ (Lubotsky and Wittenberg 2006).

Similarly, the coefficients on TA and AGE are still positive and statistically significant for the

INDIR regression. More to the point, the Wald (1943) tests–for the joint significance of the

coefficients of all the three measures of scope of operations (TA, AGE and DIVER) for BS and

INDIR regression (in Panel D of Table 4)–indicate statistically significant F-statistics. Thus,

hypothesis H1 is well-supported and confirms that bank board size and the percentage of

independent directors are positively related to the bank’s scope of operations.

With regard to the monitoring hypothesis, i.e., H2, in the BS regression, the coefficient

on the EINDEX is positive and statistically significant at the 5% level, and the coefficients on

the CV and RISK are both negative but statistically significant for CV at 10%. In the INDIR

regression, while the coefficient on EINDEX is not statistically significant, the coefficients on

CV and RISK are both negative and statistically significant at 10% or better.14 The Wald (1943)

tests for the joint significance of these two coefficients (CV and RISK) of the monitoring costs

(in Panel D of Table 4) produce statistically significant F-statistics for INDIR regression, but not

for BS regression. Thus, the overall monitoring hypothesis (H2) is partially supported, as board

size is positively related to the monitoring management’s private benefits, while the percentage

of independent directors negatively relate to directors’ monitoring costs.

With regard to the hypothesis related to CEO negotiation and succession planning (i.e.,

H3), as expected, none of the negative coefficients on all of the three measures of CEO power

(lag ROA, CEOT, and CEOWN) and a positive coefficient on the proxy for CEO succession

planning (CEOAGE) is statistically significant in the INDIR regression. Similarly, the Wald

(1943) test statistic for the joint significance of these four variables (lag ROA, CEOT, CEOWN

and CEOAGE) is not statistically significant (F-statistic = 0.40, p-value = 0.81). Contrary to the

prediction, the coefficients on the two proxies for constraints on CEO power, OUTSIDEOWN

and NBLOCKOWN, are negative and statistically significant for the former at the 1% level. The

results of OUTSIDEOWN and NBLOCKOWN, however, are consistent with findings by Linck

et al. (2008) for non-bank firms and Rediker and Seth (1995) and Whidbee (1997) for banks.

They interpret these results as the substitution effects of governance mechanisms, i.e., fewer

outside monitors are required when each director and non-blockholder has stronger incentives to

monitor. Thus, as anticipated, hypothesis H3 is not supported for banks. That is, bank board

14 In an unreported regression where RISK is the only proxy for monitoring costs, along with other relevant variables, a statistically significant negative coefficient on RISK is indicated.

18

independence neither decreases with CEO power and closeness of CEO retirement, nor

increases with constraints on such power.

With regard to the hypothesis related to ownership incentives (H4) in BS regression, as

anticipated, the coefficient on CEOWN is negative and statistically significant at the 1% level.

Likewise, the coefficient on OUTSIDEOWN is positive and statistically significant at the 1%

level. The Wald (1943) test for the joint significance of the two variables, CEOWN and

OUTSIDEOWN (in Panel D of Table 4), shows a statistically significant F-statistic of 17.20 (p-

value < 0.01). Thus, hypothesis H4 is well-supported and demonstrates that bank board size is

negatively related to insiders’ incentives alignment and positively related to outsiders’ incentives

alignment.

The results for DUAL regression, i.e. for equation (3), show that the coefficients on all

three measures of scope of bank operations, TA, AGE and DIVER, are positive and statistically

significant for TA and DIVER. The statistically significant F-statistics with Wald (1943) test (in

Panel D of Table 4) for the joint significance of TA, AGE and DIVER also confirms hypothesis

H5A that the probability of CEO duality increases with bank’s scope of operation. However, the

coefficient on EINDEX is negative, but not statistically significant. The coefficient on CV is

positive and statistically significant at the 5% level, while the coefficient on RISK is not

statistically significant. Yet, for non-bank firms, Linck et al. (2008) find no relation between CEO

duality and directors’ monitoring costs. The Wald (1943) test indicates the joint significance of

the CV and RISK, as shown in Panel D of Table 4, with an F-statistic of 1.59 (p-value < 0.10).

These findings lend partial support to hypothesis H5B that CEO duality is positively related to the

monitoring costs, but not negatively related to the monitoring of private benefits. With regard to

hypothesis H5C, the coefficients on all the four measures of both CEO power and succession

planning (lag ROA, CEOT, CEOWN and CEOAGE) are positive and statistically significant at

the 5% level or greater for CEOT, CEOWN and CEOAGE. As anticipated, the coefficients on

OUTSIDEOWN and NBLOCKOWN, the proxy for constraints on CEO power, are both

negative and statistically significant at 5% for OUTSIDEOWN. The statistically significant Wald

(1943) test statistics for the joint significance of all four variables related to CEO power and

succession planning, and constraints thereof (in Panel D of Table 4) further supports the results.

Taken together, hypothesis H5C is well-supported. That is, banks are more likely to combine both

CEO and board chair positions, when the CEO has greater power and closeness to retirement,

and constraints on such power are restricted.

19

The results for control variables yield further insights into the forces shaping bank

boards. The statistically significant positive coefficient on lag BS in INDIR regression indicates

that the percentage of independent directors increases with the prior period director numbers

(BS). Similarly, the positive and statistically significant coefficients on lag INDIR in BS

regression show that bank board size is positively related to the prior period percentage of

independent directors. The statistically significant coefficients on MERGER in BS and DUAL

regressions illustrate that any recent M&A activity (MERGER) is associated with larger boards

and the lower probability of combining both CEO and board chair roles. This occurs in case

acquisitions representatives from the target bank’s board are likely to be added to the acquirer

bank’s board, in order to attract the target to accept the bid offer and also to share the board

leadership structure. Finally, the statistically significant coefficients on DSOX in all three

regressions suggest that in the post-SOX period, bank board size decreased, the percentage of

independent directors increased, and the likelihood of banks combining both CEO and chair

titles decreased.

6. Robustness tests

As mentioned earlier in Section 4.3.2, this paper uses several additional tests to check the

robustness of the results with respect to potential endogeneity problems in some explanatory

variables, cross-sectional dependence in residuals, non-normality in residuals from outliers (if

any), and reducing attenuation bias from multiple proxies.

6.1 Results for three-stage least-squares (3SLS)

Following Agrawal and Knoeber (1996), equations (1), (2) and (3) for BS, INDIR, and

DUAL, respectively, are all estimated in a simultaneous system using 3SLS technique and the

results are reported in Table 5 below.


Column 3 of Table 5 reports the determinants for bank board size (BS), as specified by

equation (1). The findings remain the same as with those reported in Table 4 with improve

statistical significance. Column 5 of Table 5 presents the findings for the percentage of

independent directors (INDIR). The interpretation of the results also remains the same as those

in Table 4, except that the coefficient on DSOX is no longer statistically significant. The findings

for CEO duality (specified by equation (3)) in Column 7 of Table 5 also do not show any

significant deviations from the results reported in Table 4. However, the negative coefficient on

NBLOCKOWN is now statistically significant.

20

Thus, even with direct control for endogeneity using 3SLS, this study finds evidence that

the costs and benefits of the board’s monitoring and advising roles could explain differences in

bank board structure (board size, board independence, and CEO duality).

6.2 Results for two-step system generalized method of moments (GMM)

Following Wintoki, Linck and Netter (2009), Table 6 below presents Arellano and Bover

(1995) and Blundell and Bond (1998) two-step ‘system GMM’ estimates of equations (1), (2), and

(3). In the system GMM, first-differenced variables are used as instruments for the equations in

levels and the estimates are robust to unobserved heterogeneity, simultaneity and dynamic

endogeneity (if any). The ‘system GMM’ estimates are obtained using the Roodman ‘xtabond2’

module in Stata and Roodman (2006) illustrates detail estimation procedure of dynamic panel

data using ‘xtabond2’. The diagnostics tests in Panel B of Table 6 show that the model is well

fitted with statistically insignificant test statistics for both second-order autocorrelation in second

differences (Π2) and Hansen J-statistics of over-identifying restrictions. The residuals in the first

difference should be serially correlated (Π1) by way of construction but the residuals in the

second difference should not be serially correlated (Π2). Accordingly, in Panel B of Table 6, we

could see statistically significant Π1 and statistically insignificant Π2 for all equations. Likewise,

the Hansen J-statistics of over-identifying restrictions tests the null of instrument validity and the

statistically insignificant Hansen J-statistics for all equations of board structure determinants

indicate that the instruments are valid in the respective estimation. Finally, the number of

instruments (i.e. 174 for equation (1) and 82 for both equations (2) and (3)) used in the model is

less than the panel (i.e. 212) which makes the Hansen J-statistics more reliable.


The interpretation of the GMM estimates in Panel A of Table 6 remains the same as in

Table 4 with only few notable differences. In the equation (1) for BS, the coefficient on

EINDEX is no longer statistically significant. The coefficients on both RISK and

OUTSIDEOWN are not also statistically significant in equation (2) for INDIR. Likewise, in the

equation (3) for DUAL, the coefficient on CEOAGE is not statistically significant while the

positive coefficient on CV is now statistically significant. However, these differences do not

discredit the overall findings regarding the determinants of board structure in banks given bank

institutional arrangements as already reported in section 5.

21

6.3 Results for other analysis

Other robustness tests include Fama-MacBeth (1973) for cross-sectional dependence,

iteratively re-weighted least squares (IRLS) for outliers, and Prais-Winsten (1954) for

heteroskedasticity, cross-sectional dependence and first-order serial dependence.

The Fama-MacBeth (1973) estimates of regression equations (1) to (3), following the

two-step procedure described in Cochrane (2004, pp. 244-247), are robust to the

contemporaneous cross-sectional dependence. In the first step, for each year, a cross-sectional

regression model is fitted. In the second step, the final coefficients are estimated as the average

of the first step coefficient estimates. The Fama-MacBeth (1973) estimates are comparable to the

pooled-OLS estimates in Table 4, and hence not tabulated. The statistical significances of some

estimates have improved. For example, the positive coefficients on TA and AGE in INDIR

regression equation (2) are now statistically significant at 1% or better.

Following Coles et al. (2008) and Linck et al. (2008), to resolve difficulties with non-

normality in residuals from outliers, regression equations (1) to (3) are re-estimated using IRLS

procedure. IRLS is derived from Huber’s (1964) M-estimators (for maximum-likelihood) family

and is resistant to any outliers or influential observations.15 Following Kmenta (1986, pp. 318-

320), regression equations (1) to (3) are also estimated with the non-conventional Prais-Winsten

(1954) procedure, in which standard errors and the variance-covariance matrix are robust to

heteroskedasticity, cross-sectional dependence and first-order serial dependence. Generally, the

results of these two procedures confirm the findings displayed in Tables 4 and 5 with improved

statistical significance levels and for this reason are not in a table.16 The explanatory power of the

models now increases substantially, ranging from 22% to 85%.

Finally, as mentioned earlier in section 5, the use of multiple proxies to measure one

variable (such as TA, AGE and DIVER for the bank’s scope of operations) could bias the

coefficient estimates toward zero (Lubotsky and Wittenberg 2006). This is because these

different proxies could be highly correlated. Therefore, the proxies for one variable are

combined to form a single factor using principal component analysis (PCA). As such, three new

proxy variables, SCOPE, MONCOSTS and CEOPOWER, are created. SCOPE is the principal

factor covering TA, AGE and DIVER, while MONCOST addresses the information contained

in CV and RISK and CEOPOWER for lag of ROA, CEOT, and CEOWN. The findings do not

deviate with the pooled-OLS estimation of regression equations (1) and (2) and the LOGIT

15 See Birkes and Dodge (1993, pp. 98-99) for a detailed discussion of IRLS estimation procedure. 16 The results are available from the author upon request.

22

estimation of equation (3), with SCOPE, MONCOSTS and CEOPOWER & SUCCESSION in

place of multiple proxies for the banks’ scope of operations, directors’ monitoring costs, and

CEO power (respectively), and hence unreported.

6.4 Results for small, medium and large banks

Although TA is incorporated to control for bank size, it could be worth investigating

whether the results remain the same for different size groups. Hence, regression equations (1) to

(3) are re-estimated for three different size groups: small, medium and large banks. The size

groups are formed by ranking the banks into quartiles based on their total assets per year. The

first quartile banks are labeled as small, the second and third quartiles as medium, and the fourth

quartile as large.


Table 7 reports the results of equations (1), (2) and (3) for these three distinct size

groups. The findings appear to remain the same for SCOPE, MONCOSTS and CEOPOWER

across these groups, in relation to bank board structure. However, there are a few significant

differences. For example, the positive coefficient on lag INDIR in BS determinants (i.e. equation

(1)) is statistically significant for small banks. Similarly, the negative coefficient on

NBLOCKOWN in INDIR determinants (equation (2)) is statistically significant for small and

large banks, but not the same for medium banks. Non-executive directors’ shareholding appears

to lower the probability of CEO duality for small and large banks, but not for medium banks, as

indicated by the negative coefficient on OUTSIDEOWN in DUAL regression equations (3). In

addition, prior merger is associated with larger board size for medium and large banks. Taken

together, the explanatory power of the models are comparable for both small and large banks,

rather than for medium banks. In conclusion, the results for this analysis should be cautiously

interpreted due to low statistical power from the small sample size for each group, compared to

non-bank findings.

7. Impact of reforms with SOX

Several studies have examined the impact of ‘reforms with SOX’ on non-financial board

structure and firm value (see, for example, Chhaochharia and Grinstein 2007; Linck et al. 2008).

More specifically, Linck et al. (2008) show that both board size and independence increased

during the post-SOX period for non-financial firms. They also demonstrate that the results for

board structure determinants do not alter notably in the post-SOX. Similarly, for financial service

firms (banks, thrifts institutions, insurance and securities), Akhigbe and Martin (2006) find a

favorable ‘valuation effect’ of SOX for those with more independent boards and audit

23

committees, financial experts on the audit committees, increased insiders’ incentives and

institutional shareholdings. Therefore, though this is not the main focus of the paper, this section

discusses the results for some univariate tests provided to show whether there are any significant

differences in bank board structure between the pre- and post-SOX periods. In this regard, both

parametric-(paired t-test) and nonparametric tests (Wilcoxon signed rank sum test) are used. In

addition, the researcher has tested in a multivariate framework whether the determinants of bank

board structure are significantly different in the post-SOX.

Table 8 reports on the changes between mean difference tests (paired t-tests) and median

difference tests (Wilcoxon signed rank sum test) of the board structure variables between the

pre- and the post-SOX periods. As board variables seem to be less time-varying, banks in 2001

(as pre-SOX period) and 2004 (as post-SOX period) are used to test for these changes.17 The

rationale is that some banks could have adopted SOX provisions prior to 2002, while others

could have been slow to implement them. Thus, by using only 2001 and 2004 as the pre- and

post-SOX periods, respectively, the change, if any, should be most notable. The bank board size

(BS) decreased by 4% (12.93 versus 12.37), with the statistically significant paired t-statistic of -

3.326 (p-value< 0.01) and Wilcoxon statistic of -2.990 (p-value < 0.01). For banks, this decrease

in board size could reflect an increase public awareness and investors’ preference for small

boards because the bank size increase was statistically significantly over the sample period. Due

to less coordination and communication costs, smaller boards are more efficient than larger

boards (Jensen 1993; Yermack 1996). Accordingly, Wu (2000) states that the California Pension

Reinvestment Services (CalPERS) is more likely to invest in firms boards with less than 15

directors.


It is not surprising to see that the percentage of executive directors on the bank board

decreased by 9%, which is statistically significant at the 1% level (paired t-statistic = -3.094 and

Wilcoxon statistic = -3.436). Similarly, the percentage of independent bank directors (INDIR)

increased by 5%, from 63.40% in 2001 to 66.59% in 2004 and is statistically significant with the

paired t-statistic of 4.665 (p-value < .01) and Wilcoxon statistic of 4.263 (p-value < 0.01).

However, the percentage of banks that combined both CEO and board chairman titles (DUAL)

decreased by 4%, from 57% in 2001 to 55% in 2004, but this change is not statistically

significant. In contrast, Linck et al. (2008) reported a statistically significant decrease in the

17 Similarly, Linck et al. (2008) also considered 2001 and 2004 as pre- and post-SOX periods, respectively, in their

study of the impact of SOX on the non-bank firms board structure.

24

percentage of non-bank firms with a combined CEO and board chair titles, from 62% in 2001 to

56% in 2004.


Table 9 presents changes in the results for the bank board structure determinants

between the post-SOX and pre-SOX periods. In regression equations (1) to (3), the post-SOX

dummy, DSOX, interacts with each of the hypothesized determinants. The results show only

three notable differences. The coefficients on both EINDEX and its interaction term with

DSOX are positive and statistically significant in the BS regression equation (1). The Wald (1943)

test for the joint significance of EINDEX and its interaction term with the DSOX also provide

statistically significant F-statistics. This indicates that EINDEX is associated with larger boards

in both periods, but the relation is stronger post-SOX, compared to pre-SOX. Finally, the

coefficient on CEOAGE is insignificant, but its interaction terms with DSOX in the INDIR

regression equation (2) is negative and statistically significant. The Wald (1943) test for the joint

significance of CEOAGE and CEOAGE*DSOX provides statistically insignificant F-statistics.

This suggests that CEOAGE do not affect board independence in both periods, but appear to

have some negative impact on INDIR post-SOX. Finally, the results for CEOPOWER and its

interaction term in DUAL regression equation (3) indicate that CEOPOWER positively affect

bank CEO duality, and this relation is even stronger in the post-SOX. Taken together, there is

some evidence that bank board structure changed in the post-SOX compared to the pre-SOX.

8. Conclusion

This paper investigates the trends and the endogenously chosen bank boards of directors,

with specific reference to the costs and benefits associated with boards’ monitoring and advising

functions. In particular, the study endeavors to assess whether bank board structure (board size,

composition and CEO duality) is associated with the bank scope of operation, trade-off between

costs and benefits of monitoring, negotiations with the bank CEO and incentives alignments,

after controlling for other bank specific characteristics.

Using a sample of 212 BHCs over the 1997-2004 period, or 1,534 bank year

observations, this study finds evidence that board structure, even in a regulated industry like

banking can be explained by its monitoring and advising roles. Consistent with expectations and

existing non-bank findings, the results demonstrate that larger and more diversified banks have

larger boards, more independent directors and combined leadership structure. In the presence of

managers’ opportunities to consume private benefits, the study indicates that banks benefit from

larger boards, while banks benefit from more independent directors when the costs of

25

monitoring managers are low. In contrast to non-bank evidence, it is argued that bank

managers–including CEOs–do not influence board selection processes due to constant

monitoring by bank regulators and fear of severe disciplinary action. Consistent with this view,

board independence is found not to be the outcome of negotiations with bank CEOs. This study

also confirms that banks have smaller boards when insiders’ shareholdings are high and the

outsiders’ shareholdings are low.

The trend analysis shows that bank boards become smaller over the sample period,

specifically for large and medium banks in the post-SOX period. Small bank board size remains

stable over this sample period. The percentage of independent directors increases remarkably for

small banks during the early 2000s. Moreover, there is some evidence that bank board structure

changed significantly in the post-SOX period. For example, the mean percentage of independent

directors increased with statistical significance in the post-SOX period, rather than in the pre-

SOX period. The average post-SOX bank board size is smaller than that of the pre-SOX bank

board.

Taken as a whole, this study provides some evidence that even in a regulatory industry

like banking, the structure of boards is consistent with the efficiency argument, i.e., banks

structure their boards in a way to maximize shareholders’ wealth. Yet the results for control

variables that board size and independence do not relate to past performance could support the

inefficiency argument put forth by Boone et al. (2007). Therefore, further study on board

structure determinants is warranted to enhance academic understanding of this subject. For bank

regulators, the findings have important policy implications. For instance, to the extent that bank

board structure adapts to its unique competitive environment, it suggests that uniform rules or

guidelines to reform board governance could prove counter-productive. Thus, bank regulators

should cautiously evaluate the effectiveness of SOX and other requirements for banks.

26

References

Adams, R. B., and Ferreira, D. (2007). A theory of friendly board. The Journal of Finance, 62(1), 217-250.

Adams, R. B., and Mehran, H. (2003). Is corporate governance different for bank holding companies. FRBNY Economic Policy Review, 9(1), 123-142.

Adams, R. B., and Mehran, H. (2009). Corporate performance, board structure and their determinants in the banking industry. Journal of Financial Intermediation, forthcoming.

Agrawal, A., and Knoeber, C. R. (1996). Firm performance and mechanisms to control agency problems between managers and shareholders. Journal of Financial & Quantitative Analysis, 31(3), 377-397.

Akhigbe, A., and Martin, A. D. (2006). Valuation impact of Sarbanes-Oxley: Evidence from disclosure and governance within the financial services industry. Journal of Banking & Finance, 30(3), 989-1006.

Alchian, A. (1950). Uncertainty, evolution and economic theory. Journal of Political Economy, 58(2), 211–221.

Anderson, R. C., and Fraser, D. R. (2000). Corporate control, bank risk taking, and the health of the banking industry. Journal of Banking & Finance, 24(8), 1383-1398.

Arellano, M., and Bover, O. (1995). Another look at the instrumental variables estimation of error-components models. Journal of Econometrics, 68(1), 29-51.

Baker, M., and Gompers, P. A. (2003). The determinants of board structure at the initial public offering. Journal of Law and Economics, 46(3), 569-598.

Bebchuk, L. A., Cohen, A., and Ferrell, A. (2009). What matters in corporate governance? Review of Financial Studies, 22(2), 783-827.

Bhagat, S., and Black, B. (2002). The non-correlation between board independence and long-term firm performance. Journal of Corporation Law, 27(2), 231-273.

Birkes, D., and Dodge, Y. (1993). Alternative methods of regression. New York: John Wiley & Sons.

Blundell, R., and Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics, 87(2), 115-143.

Boone, A. L., Field, C. L., Karpoff, J. M., and Raheja, C. G. (2007). The determinants of corporate board size and composition: an empirical analysis. Journal of Financial Economics, 85(1), 66-101.

Booth, J. R., Cornett, M. M., and Tehranian, H. (2002). Boards of directors, ownership, and regulation. Journal of Banking & Finance, 26(10), 1973-1996.

Brickley, J. A., Coles, J. L., and Jarrell, G. (1997). Leadership structure: separating the CEO and chairman of the board. Journal of Corporate Finance, 3(3), 189-220.

Brickley, J. A., Coles, J. L., and Terry, R. L. (1994). The board of directors and the enactment of poison pills. Journal of Financial Economics, 35(3), 371-390.

Coles, J. L., Daniel, N. D., and Naveen, L. (2008). Boards: does one size fit all? Journal of Financial Economics, 87(2), 329-356.

Cornett, M. M., McNutt, J., and Tehranian, H. (2009). Corporate governance and earnings management at large U.S. bank holding companies. Journal of Corporate Finance, 15(2), 412-430.

27

Demirgüç-Kunt, A., and Detragiache, E. (2002). Does deposit insurance increase banking system stability? An empirical investigation. Journal of Monetary Economics, 49(7), 1373-1406.

Fama, E. F., and Jensen, M. C. (1983). Agency problems and residual claims. Journal of Law and Economics, 26(2), 327-342.

Fama, E. F., and MacBeth, J. (1973). Risk, return and equilibrium: empirical tests. Journal of Political Economy, 71(3), 607-636.

Fanto, J. A. (2006). Paternalistic regulation of public company management: lessons from bank regulation. Brooklyn Law School Legal Studies Paper No. 49. Available at SSRN <http://ssrn.com/abstract=873667>, viewed on August 29, 2009.

Fisher, D. B. (1992). Bank director liability under FIRREA: a new defense for directors and officers of insolvent depository institutions or a tighter noose. UCLA Law Review, August, 1703-1790.

Flannery, M. J. (1998). Using market information in prudential bank supervision: A review of the U.S. empirical evidence. Journal of Money, Credit and Banking, 30(3), 273-305.

Flannery, M. J., and Sorescu, S. M. (1996). Evidence of bank market discipline in subordinated debenture yields. The Journal of Finance, 51(4), 1347-1377.

Galai, D., and Masulis, R. W. (1976). The option pricing model and the risk factor of stock. Journal of Financial Economics, 3(1-2), 53-81.

Gillan, S. L., Hartzell, J. C., and Starks, L. T. (2006). Tradeoffs in corporate governance: Evidence from board structures and charter provisions. Working Paper No. n/a, University of Texas at Austin, Austin.

Harris, M., and Raviv, A. (2008). A theory of board control and size. Review of Financial Studies, 21(4), 1797-1832.

Hermalin, B. E., and Weisbach, M. S. (1998). Endogenously chosen boards of directors and their monitoring of the CEO. The American Economic Review, 88(1), 96-118.

Hermalin, B. E., and Weisbach, M. S. (2003). Boards of directors as an endogenously determined institution: A survey of the economic literature. FRBNY Economic Policy Review, 9(1), 7-26.

Huber, P. J. (1964). Robust estimation of a location parameter. The Annals of Mathematical Statistics, 35(1), 73-101.

Jensen, M. C. (1993). The modern industrial revolution, exit, and the failure of internal control systems. The Journal of Finance, 48(3), 831-880.

Keeley, M. C. (1990). Deposit insurance, risk and market power in banking. The American Economic Review, 80(5), 1183-1200.

Kmenta, J. (1986). Elements of econometrics (2nd ed.). New York: Macmillan

Lehn, K., Patro, S., and Zhao, M. (2009). Determinants of the size and structure of corporate boards: 1935-2000. Financial Management, forthcoming.

Linck, J. S., Netter, J. M., and Yang, T. (2008). A large sample study on board changes and determinants of board structure. Journal of Financial Economics, 87(2), 308-328.

Lubotsky, D., and Wittenberg, M. (2006). Interpretation of regressions with multiple proxies. Review of Economics and Statistics, 88(3), 549-562.

Mace, M. L. (1971). Directors, myth and reality. Boston: Harvard Business School Press.

28

29

Macey, J. R. and Miller, J. P. (1993). Corporate stakeholders: a contractual perspective. The University of Toronto Law Journal, 43(3), 401-424.

Macey, J. R., and O'Hara, M. (2003). The corporate governance of banks. FRBNY Economic Policy Review, 9(1), 91-107.

Marcus, A. J. (1984). Deregulation and bank financial policy. Journal of Banking & Finance, 8(4), 557-565.

McConnell, J. J., and Servaes, H. (1990). Additional evidence on equity ownership and corporate value. Journal of Financial Economics, 27(2), 595-612.

Morck, R., Shleifer, A., and Vishny, R. W. (1988). Managerial ownership and market valuation: An empirical analysis. Journal of Financial Economics, 20(4), 293-315.

Petersen, M. A. (2009). Estimating standard errors in finance panel data sets: comparing approaches. Review of Financial Studies, 22(1), 435-480.

Raheja, C. G. (2005). Determinants of board size and composition: A theory of corporate boards. Journal of Financial & Quantitative Analysis, 40(2), 283-306.

Rediker, K. J., and Seth, A. (1995). Boards of directors and substitution effects of alternative governance mechanisms. Strategic Management Journal, 16(2), 85-99.

Roodman, D. (2006). How to do xtabond2: An introduction to "difference" and "system" GMM in Stata. Working paper, Center for Global Development, Washington.

Shepherd, J. (1992). The liability of officers and directors under the Financial Institutions Reform, Recovery and Enforcement Act of 1989. Michigan Law Review, 90(5), 1119-1144.

Stiroh, K. J., and Rumble, A. (2006). The dark side of diversification: The case of US financial holding companies. Journal of Banking & Finance, 30(8), 2131-2161.

Vancil, R. F. (1987). Passing the baton: managing the process of CEO succession. Boston: Harvard business School Press.

Wintoki, M. B., Linck, J. S., and Netter, J. M. (2009). Endogeneity and the dynamics of corporate governance. Available at SSRN <http://ssrn.com/abstract=970986>, viewed on October 31, 2009.

Whidbee, D. A. (1997). Board composition and control of shareholder voting rights in the banking industry. Financial Management, 26(1), 27-41.

White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48(4), 817-838.

Wooldridge, J. M. (2002). Econometric analysis of cross section and panel data. Cambridge, Massachusetts; London: MIT Press.

Wooldridge, J. M. (2006). Introductory econometrics: A modern approach (3rd ed.). Mason, OH: Thomson/South Western.

Wu, Y. (2000). Honey, CalPERS shrunk the board. Ph.D. Thesis, University of Chicago, Chicago.

Yermack, D. (1996). Higher market valuation of companies with a small board of directors. Journal of Financial Economics, 40(2), 185-212.

Panel A: Board Size

10.00

11.00

12.00

13.00

14.00

15.00

16.00

17.00

1997 1998 1999 2000 2001 2002 2003 2004

Small Medium Large

Panel B: % Independent Directors

50.00

55.00

60.00

65.00

70.00

75.00

1997 1998 1999 2000 2001 2002 2003 2004

Small Medium Large

Panel C: % of Banks with CEO is also Board Chair

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

0.90

1997 1998 1999 2000 2001 2002 2003 2004

Small Medium Large

Figure 1: Bank Board Structure Trends: 1997-2004. The sample includes 212 BHCs over 1997-2004, i.e. a total bank observations of 1,534. The size groups are formed by ranking the banks into quartiles based on their total assets each year. We label the first quartile banks “small”, quartiles second and third “medium” and the fourth quartile “large”. Panel A, B, and C report the trends in the mean board size, the percentage of independent directors, and the percentage of banks with CEO is also board chair, respectively for banks.

30

Table 1: Summary of Definitions and Predicted Signs of Variables for Equations (1) to (3) This table summarizes the definitions and predicted signs of the variables in regression Equations (1) to (3) related to the testing of hypotheses H1, H2, H3, H4, and H5A & 5B. Column 1 shows the list of variables and their definitions are in column 2. Columns 3 to 6 show coefficients along with their predicted signs in regression Equations (1) to (3) for the determinants of respective BHC board size (BS), percentage of independent directors (INDIR), and CEO duality (DUAL).

Variables Definitions BS INDIR DUAL

Panel A: Dependent variables Equation (1) (2) (3) BS The number of directors in the BHC’s board. INDIR The percentage of independent directors in the BHC’s

board.

DUAL A dummy variable that equals 1 if the CEO also chairs the BHC’s board, otherwise zero.

Panel B: Explanatory variables:

TA The total assets of the BHC as at the end of each fiscal year.

β1 (+) β1 (+)

AGE The number of years since the BHC was listed in the DATASTREAM database.

β2 (+) β2 (+)

DIVER Stiroh and Rumble’s (2006) ‘revenue diversification index’ which is calculated as 1 – (squared of fraction of operating income from interest plus squared of fraction of net operating income from non-interest sources).

β3 (+) β3 (+)

EINDEX The sum of two dummy variables: staggered boards, and poison pills. The dummy for staggered boards equals 1 if the BHC’s board is classified, otherwise zero. The dummy for poison pills equals one if the BHC’s board has the poison pill provision, otherwise zero.

δ1 (+) δ1 (+) δ1 (–)

CV Keeley's Q (Keeley 1990) which is calculated as the sum of the market value of equity plus the book value of liabilities divided by the book value of total assets.

δ2 (–) δ2 (–) δ2 (+)

RISK The standard deviation of daily BHC’s stock returns in a year

δ3 (–) δ3 (–) δ3 (+)

ROA The return on average total assets which is calculated as net income after tax as a percentage of the BHC’s average total assets. Average total assets is simply the average of beginning and end of year BHC’s total assets.

φ1 (–) φ1 (+)

CEOT The number of years the BHC CEO has served in this position.

φ2(–) φ2 (+)

CEOAGE The BHC CEO's age in years. φ3(–) φ3 (+) CEOWN The percentage of total outstanding shares owned by

the BHC CEO. γ1 (–) φ4 (–) φ4(+)

OUTSIDEOWN The percentage of total outstanding shares owned by the BHC officers and directors excluding those of the CEO.

γ2 (+) λ1 (+) λ1 (–)

NBLOCKOWN The percentage of total outstanding shares owned by non-affiliated block-holders who hold at least 5% of outstanding shares.

λ2 (+) λ2 (–)

Panel C: Other control variables: CAPITAL The BHC's total equity as a percentage of total assets. ζ2 (+) ζ2 (+) ζ2 (–)MERGER A dummy for any M&A, i.e. a dummy variable which

equals one for BHC that made an acquisition in a year, otherwise zero.

ζ3 (+) ζ3 (+) ζ3 (–)

DSOX A dummy for post-SOX periods, 2003 and 2004, i.e. a dummy variable which equals one if the period is either 2003 or 2004, otherwise zero.

φ (+/-) φ (+/-) φ (+/-)

31

32

Table 2: Descriptive Statistics This table presents the distribution of variables by showing mean, standard deviation (SD), minimum (Min.), first quartile (1st Quartile), median (Median), second quartile (2nd Quartile), skewness (Skew.), and kurtosis (Kurt.). See Tables 1 for variable definitions.

Variables Mean SD Min. 1st

Quartile Median2nd

Quartile Max. Skew. Kurt.

Panel A: Board structure variables: BS (No.) 12.92 4.54 5 10 12 15 31 0.96 3.83OUTDIR (%) 84.62 8.73 37.5 80 86.96 90.91 100 -1.49 5.94INDIR (%) 64.52 15.72 10 55.56 66.67 75 96.55 -0.58 3.1STAGG 0.74 0.44 0 0 1 1 1 -1.11 2.22POISON 0.34 0.47 0 0 0 1 1 0.7 1.49EINDEX 1.08 0.72 0 1 1 2 2 -0.12 1.91Panel B: CEO characteristics: DUAL 0.58 0.49 0 0 1 1 1 -0.31 1.1CEOT (Years) 8.85 7.74 0 3 7 14 46 1.18 4.52CEOAGE (Years) 56.09 7.32 34 51 56 60 85 0.4 3.84CEOWN (%) 4.41 8.8 0 0.55 1.3 3.46 65.19 3.72 18.72Panel C: Ownership structure: OUTSIDEOWN (%) 10.25 9.96 0.19 4.04 7.24 13.36 83.32 2.7 14.46NBLOCKOWN (%) 3.67 8.54 0 0 0 5.7 98.5 4.9 38.07Panel D: Bank-specific variables: TA (in bil.) 23.66 105.78 .16 1.02 2.07 7.66 1484.10 8.23 81.31CV 1.1 0.07 0.94 1.05 1.09 1.13 1.64 1.82 10.54CAPITAL (%) 9.26 1.9 3 7.99 9.09 10.16 21.59 1.34 7.93DIVER 0.36 0.09 0.06 0.3 0.36 0.43 0.5 -0.53 2.71AGE (Years) 13.73 9.01 1 6 12 22 31 0.46 1.9MERGER 0.11 0.32 0 0 0 0 1 2.45 6.99RISK (%) 2.26 1.2 0.65 1.65 2.02 2.53 17.32 4.81 42.75ROA(%) 1.24 0.51 -6.24 1 1.22 1.47 5.59 -1.01 41.62

Table 3: Correlation Matrix This table shows Pearson pair-wise correlation matrix. Bold texts indicate statistically significant at 1% level or better. See Table 1 for variables definitions.

Variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 171 BS 1.00 0.16 0.01 0.08 0.00 -0.01 -0.15 0.08 -0.04 0.38 0.02 -0.02 0.28 0.18 0.05 -0.1 -0.052 INDIR 1.00 0.05 0.14 -0.05 0.01 -0.12 -0.36 0.04 0.33 0.05 0.01 0.17 0.32 0.00 -0.18 0.023 EINDEX 1.00 -0.04 -0.18 -0.17 -0.14 -0.25 -0.05 -0.01 0.03 -0.07 -0.02 0.05 -0.02 -0.12 -0.034 DUALCEO 1.00 0.25 0.22 0.11 -0.19 -0.03 0.24 0.09 -0.07 0.20 0.22 -0.07 -0.04 0.055 CEOT 1.00 0.42 0.18 0.08 -0.06 -0.04 -0.03 0.04 -0.01 0.05 -0.05 -0.02 0.006 CEOAGE 1.00 0.03 0.02 -0.04 0.01 0.01 0.08 -0.05 0.08 -0.02 0.04 0.057 CEOWN 1.00 0.06 0.04 -0.15 -0.13 0.02 0.00 -0.10 -0.07 0.09 0.008 OUTSIDEOWN 1.00 -0.10 -0.34 -0.18 -0.13 -0.19 -0.37 -0.02 0.08 -0.14

9 NBLOCKOWN 1.00 0.07 0.01 -0.01 0.02 0.09 -0.04 -0.03 0.0410 LNTA 1.00 0.31 -0.07 0.51 0.68 0.12 -0.24 0.20

11 CV 1.00 0.17 0.08 0.29 -0.02 -0.13 0.59

12 CAPITAL 1.00 -0.07 0.08 0.08 0.01 0.40

13 DIVER 1.00 0.39 0.01 -0.18 0.0614 AGE 1.00 0.00 -0.17 0.23

15 MERGER 1.00 -0.03 -0.0416 RISK 1.00 -0.19

17 ROA 1.00

33

Table 4: Regression Results for the Determinants of the Bank Board Structure This table presents the results of the pooled ordinary least squares (OLS) estimates of equations (1) and (2) and LOGIT estimates of equation (3). Standard errors in all estimations are clustered by banks. BS is the number of directors on the board. INDIR is the independent directors as a percentage of board size. DUAL is the dummy variable which equals 1 if the CEO also chairs the board. TA is the total assets at fiscal year-end. AGE is the number of years since the BHC was listed in the DATASTREAM database. DIVER is the revenue diversification index calculated following Stiroh and Rumble (2006). EINDEX is the sum of the two variables - staggered board and poison pill, and is an approximation of Bebchuk et al. (2009) entrenchment index. CV is the charter value of the bank calculated (following Keeley (1990)) as the book value of total assets plus market value of equity minus book value of equity, all divided by the book value of total assets. RISK is the standard deviation of the bank’s daily stock returns over a year. ROA is the net income after tax as a percentage of average total assets. CEOT is the number of years the CEO has held this position. CEOWN is the percentage of shares owned by the CEO. CEOAGE is the age of the CEO in years. OUTSIDEOWN is the percentage of shares owned by the directors and top executives excluding CEOWN. NBLOCKOWN is the percentage of shares owned by non-affiliated persons/institutions with 5% or more of the bank’s equity. CAPITAL is the bank equity as percentage of total assets. MERGER is the dummy variable which equals 1 if the bank has any M&A in the period. DSOX is the dummy variable which equals 1 if the period is either 2003 or 2004. AVIF is the average ‘variance inflation factor’ shows the degree of collinearity problem among the regressors. Π1 is the White (1980) test for heteroskedasticity which provides the Lagrange Multiplier (LM) statistics based on alternative procedure explained in Wooldridge (2006, pp. 282-283). Π2 is the test for first-order serial correlation which provides an F-statistic based on Wooldridge (2002, pp. 282-283). Finally, the Wald (1943) test is used to assess the joint significance of the respective estimates. Figures in parentheses show the robust t-statistics based on standard errors clustered by both bank and year. Figures in brackets present the p-values of the respective F-statistics from the Wald (1943) tests. Superscripts *, **, *** indicate statistical significance at 10%, 5%, and 1% levels, respectively. Variables Pred. BS (1) Pred. INDIR (2) Pred. DUAL (3) Panel A: Explanatory variables: TA + 0.087***(6.74) + 0.021*(1.72) + 0.225*(1.68)AGE + -0.048(-1.61) + 0.033*(1.67) + 0.162(0.63)DIVER + 0.596***(3.36) + -0.115(-0.62) + 2.568*(1.68)EINDEX + 0.05**(2.09 + -0.016(-0.74) - -0.039(-0.18)CV - -0.367*(-1.68) - -0.350**(-2.02) + 1.892**(1.97)RISK - -0.005(-0.50) - -0.034*(-1.75) + 0.0413(0.47)ROAt-1 - -0.008(-0.29) + -0.127(-0.70)CEOT - -0.001(-0.58) + 0.069***(3.10)CEOWN - -0.004***(-2.59) - -0.001(-0.72) + 0.035**(1.75)CEOAGE - 0.037(0.30) + 2.416**(2.23)OUTSIDEOWN + 0.009***(4.53) + -0.009***(-5.09) - -0.0339**(-2.53)NBLOCKOWN + 0.0003(0.14) - -0.016(-1.16)BS t-1 + 0.123***(2.77) INDIR t-1 - 0.167***(3.11) + 0.213(0.48)CAPITAL + 0.013*(1.72) + 0.0009(0.14) - -0.144**(-2.16)MERGERt-1 + 0.095***(2.96) + 0.007(0.37) - -0.487**(-2.25)DSOX +/- -0.06***(-3.62) + 0.009*(1.78) +/- -0.301**(-2.32)Constant 1.164***(2.97) 4.042***(8.2) -14.19***(-2.80)Panel B: Model fits Adjusted R2/Pseudo R2 0.2755 0.2125 0.1674F-stats. (13|16|17, 1309) 55.57***[0.00] 19.81***[0.00] 29.84***[0.00]No. of pooled obs. 1322 1322 1322Panel C: Regression diagnostics: AVIF (max.) 1.34 (2.25) 1.41 (2.50) 1.39 (2.28)Π1: LM-stats (χ2 = 2) 13.33*** [0.00] 89.03*** [0.00] 85.71*** [0.00]Π2: F-stats (1, 211) 47.94*** [0.00] 29.75*** [0.00] 74.61*** [0.00]Panel D: Wald test (F stat) for joint significance of

SCOPE, F (3, 1309|1305) 33.72***[0.00] 3.15** [0.02] 6.90*** [0.00]MONCOST, F (2,

1309|1305)

1.63 [0.20] 2.89* [0.06]

1.59* [0.07]INCENTIVE, F (2, 1309) 17.20*** [0.00] N/A N/A

CEOPOWER & SUCCESSION, F (4, 1305)

N/A .40 [0.81]

6.86*** [0.00]

Constraints on CEO power, F (2, 1305)

N/A 13.03*** [0.00]

3.67** [0.03]

34

35

Table 5: 3SLS Regression Results for the Determinants of the Bank Board Structure This table presents the results of the three-stage least squares (3SLS) estimation of the system of simultaneous equations (1) to (3). BS is the number of directors on the board. INDIR is the independent directors as a percentage of board size. DUAL is the dummy variable which equals 1 if the CEO also chairs the board. TA is the total assets at fiscal year-end. AGE is the number of years since the BHC was listed in the DATASTREAM database. DIVER is the revenue diversification index calculated following Stiroh and Rumble (2006). EINDEX is the sum of the two variables - staggered board and poison pill, and is an approximation of Bebchuk et al. (2009) entrenchment index. CV is the charter value of the bank calculated (following Keeley (1990)) as the book value of total assets plus market value of equity minus book value of equity, all divided by the book value of total assets. RISK is the standard deviation of the bank’s daily stock returns over a year. ROA is the net income after tax as a percentage of average total assets. CEOT is the number of years the CEO has held this position. CEOWN is the percentage of shares owned by the CEO. CEOAGE is the age of the CEO in years. OUTSIDEOWN is the percentage of shares owned by the directors and top executives excluding CEOWN. NBLOCKOWN is the percentage of shares owned by non-affiliated persons/institutions with 5% or more of the bank’s equity. CAPITAL is the bank equity as percentage of total assets. MERGER is the dummy variable which equals 1 if the bank has any M&A in the period. DSOX is the dummy variable which equals 1 if the period is either 2003 or 2004. Finally, the Wald (1943) test is used to assess the joint significance of the respective estimates. Figures in parentheses show the t-statistics while those in brackets presents the p-values of the respective F-statistics from the Wald (1943) tests. Superscripts *, **, *** indicate statistical significance at 10%, 5%, and 1% levels, respectively.

Variables Pred. sign BS (eq.1)

Pred. sign INDIR (eq.2)

Pred. sign DUAL (eq.3)

Panel A: Explanatory variables: TA + 0.084***(11.13) + 0.014**(1.99) + 0.0430***(3.7)AGE + -0.052(-1.11) + 0.035***(2.70) + 0.033(1.47)DIVER + 0.609***(6.03) + -0.161(1.56) + 0.494***(3.17)EINDEX + 0.052***(4.42) + -0.0187(-.76) - 0.0006(0.03)CV - -0.319***(2.64) - -0.303**(-2.46) + 0.365*(1.71)RISK - 0.008(1.13) - -0.0329***(-5.07) + 0.007(0.63)ROAt-1 - -0.009(-0.51) + -0.005(-0.15)CEOT - -0.001(-0.95) + 0.0128***(7.14)CEOWN - -0.004***(-4.43) - -0.001(-1.24) + 0.006***(4.32)CEOAGE - 0.035(0.57) + 0.471***(4.41)OUTSIDEOWN + 0.01***(10.13) + -0.009***(-11.08) - -0.007***(-4.63)NBLOCKOWN + -0.0003(-0.33) - -0.003**(-2.29)BS t-1 + 0.202***(8.57) INDIR t-1 - 0.26***(8.99) + 0.047(1.06)CAPITAL + 0.012***(2.85) + 0.0002(0.05) - -0.029***(-4.04)MERGERt-1 + 0.096***(3.62) + -0.008(-0.33) - -0.10**(-2.51)DSOX +/- -0.06***(-3.22) + 0.0112(0.68) +/- -0.06**(-2.14)Constant 0.741*(3.88) 3.883***(13.63) -2.295***(-4.42)Panel B: Model fits Adjusted R2/Pseudo R2 0.2598 0.2059 0.2037χ2-stats. (13|16|16, 1534) 525.55***[0.00] 403.48***[0.00] 337.35***[0.00]No. of banks 212 212 212No. of pooled obs. 1532 1532 1532Panel C: Wald test (F stat) for joint significance of

SCOPE, F (4|3|3, 211) 90.12***[0.00] 6.86*** [0.00] 22.37*** [0.00]MONCOST, F (2, 211) 4.25*** [0.01] 16.24*** [0.00] 2.74*[0.07]

INCENTIVE, F (2, 211) 63.78*** [0.00] N/A N/ACEOPOWER &

SUCCESSION, F (4, 211)

N/A 3.04 [0.17]

49.72*** [0.00]Constraints on CEO power,

F (2, 211)

N/A 62.36*** [0.00]

12.45*** [0.00]

Table 6: Two-Step System GMM Regression Results for the Determinants of the Bank Board Structure

This table presents the results of the two-step system GMM (OLS) estimates of equations (1), (2), and (3). BS is the number of directors on the board. INDIR is the independent directors as a percentage of board size. DUAL is the dummy variable which equals 1 if the CEO also chairs the board. TA is the total assets at fiscal year-end. AGE is the number of years since the BHC was listed in the DATASTREAM database. DIVER is the revenue diversification index calculated following Stiroh and Rumble (2006). EINDEX is the sum of the two variables - staggered board and poison pill, and is an approximation of Bebchuk et al. (2009) entrenchment index. CV is the charter value of the bank calculated (following Keeley (1990)) as the book value of total assets plus market value of equity minus book value of equity, all divided by the book value of total assets. RISK is the standard deviation of the bank’s daily stock returns over a year. ROA is the net income after tax as a percentage of average total assets. CEOT is the number of years the CEO has held this position. CEOWN is the percentage of shares owned by the CEO. CEOAGE is the age of the CEO in years. OUTSIDEOWN is the percentage of shares owned by the directors and top executives excluding CEOWN. NBLOCKOWN is the percentage of shares owned by non-affiliated persons/institutions with 5% or more of the bank’s equity. CAPITAL is the bank equity as percentage of total assets. MERGER is the dummy variable which equals 1 if the bank has any M&A in the period. DSOX is the dummy variable which equals 1 if the period is either 2003 or 2004. Finally, Π1 and Π2 are the test statistics for first-order and second-order serial correlation respectively. Hansen J-statistics is the test of over-identifying restrictions. Figures in parentheses are t-statistics while p-values are in brackets. Superscripts *, **, *** indicate statistical significance at 10%, 5%, and 1% levels, respectively.

Variables Pred. sign BS (1)

Pred. sign INDIR (2)

Pred. sign DUAL (3)

Panel A: Explanatory variables: TA + 0.084***(4.39) + 0.024*(1.84) 0.069**(2.41)AGE + -0.052(-1.24) + 0.053*(1.86) -0.001(-0.04)DIVER + 0.48***(2.67) + -0.084(-0.43) 0.53*(1.72)EINDEX + 0.025(0.75) + -0.011(-0.42) - -0.006(-0.10)CV - -0.84**(-2.01) - -0.553*(-1.70) + 0.191*(1.68)RISK - 0.006(0.52) - -0.02(-0.80) + 0.013(0.54)ROAt-1 - -0.001(-0.03) + 0.049(1.31)CEOT - 0.001(0.04) + 0.0185***(2.61)CEOWN - -0.007*(-1.89) - -0.0003(-0.07) + 0.001*(-1.67)CEOAGE + -0.168(-0.60) - 0.486(1.09)OUTSIDEOWN + 0.003*(1.72) + -0.006(-1.41) - -0.009(-1.26)NBLOCKOWN + -0.003(-0.81) -0.007(-0.75)BS t-1 + 0.033*(1.86) - INDIR t-1 - 0.117*(1.69) - + -0.059(-0.37)CAPITAL + -0.009(-0.67) + -0.008(-0.77) - -0.025*(-1.79)MERGERt-1 + 0.068***(4.41) + -0.002(-0.12) - -0.0457*(-1.69)DSOX +/- -0.07**(-1.98) 0.007*(1.72) -0.153***(-3.47)Constant 2.28***(3.23) 5.27***(3.90) -1.8984(-0.93)Year dummies Included Included IncludedPanel B: Model fits F-statistics (16|20|20, 211) 8.72 [0.00] 2.61 [0.00] 3.99 [0.00]Π1 -3.40***[0.00] -2.37**[0.018] -3.15***[0.00]Π2 -0.22 [0.829] -0.54 [0.586] -1.03 [0.304]Hansen J-statistics (χ2=156|60|60)

168.20 [0.239] 43.89 [0.941] 58.57 [0.528]

No. of instruments 174 82 82No. of banks 212 212 212No. of pooled observations 1322 1322 1322

36

Table 7: Determinants of the Board Structure for Small, Medium and Large Banks This table presents the results of the pooled ordinary least squares (OLS) estimates of equations (1) and (2) and LOGIT estimates of equation (3). Standard errors in all estimations are clustered by banks. BS is the number of directors on the board. INDIR is the independent directors as a percentage of board size. DUAL is the dummy variable which equals 1 if the CEO also chairs the board. SCOPE is the PCA factor covering TA, AGE and DIVER. MONCOST is the PCA factor for CV and RISK. CEOPOWER is the PCA factor for lag of ROA, CEOT, and CEOWN. CEOAGE is the age of the CEO in years. OUTSIDEOWN is the percentage of shares owned by the directors and top executives excluding CEOWN. NBLOCKOWN is the percentage of shares owned by non-affiliated persons/institutions with 5% or more of the bank’s equity. CAPITAL is the bank equity as percentage of total assets. MERGER is the dummy variable which equals 1 if the bank has any M&A in the period. DSOX is the dummy variable which equals 1 if the period is either 2003 or 2004. Figures in parentheses show the robust t-statistics while those in brackets presents the p-values of the respective F-statistics from the Wald (1943) tests. Superscripts *, **, *** indicate statistical significance at 10%, 5%, and 1% levels, respectively.

Pred BS (eq.1) Pred INDIR (eq.2)

Variables sign Small Medium Large sign Small Medium Large

SCOPE + .054* .088*** .07* + -0.086* .058** .037* (1.69) (2.70) -1.74 (-1.87) (2.03) (1.87)EINDEX + -.013 .049* .06* + 0.057 -.026 -.024 (-1.76) -1.68 -1.74 (1.16) (-0.85) (-0.93)MONCOSTS - -.003 .017 .049 - -0.04** -.013* .014* (-0.11) (0.62) (0.98) (-1.96) (-1.75) (1.67)CEOPOWER - -0.09 -.004 -.023 (-1.58) (-0.15) (-0.98)CEOAGE - 1.02 .13 .12 (0.82) (0.56) (0.55)CEOWN - -.004* -.005* -.011* (-1.77) (-1.84) (-1.74) OUTSIDEOWN + .009*** .007* .011*** + -0.014*** -.007** -.007*** (2.84) (1.89) (4.07) (-3.53) (-2.32) (-5.80)NBLOCKOWN + 0.013*** .003 -.004*** (4.28) (1.47) (-3.27)BS t-1 + 0.51*** .049 .091 (3.46) (0.87) (1.16)INDIR t-1 +- .28*** .073 .123 (3.91) (0.81) (0.63) CAPITAL + -.006 .015 .011 + -0.002 .005 -.003 (-0.39) (0.94) (0.48) (-0.12) (0.49) (-0.54)MERGERt-1 + .084 .13*** .15*** + 0.021 .034 .01 (1.57) (4.00) (3.02) (0.38) (1.08) (0.27)DSOX +/- -.07 -.07 -.05 + -0.0003 -.0189 .017 (-0.98) (-1.37) (-1.30) (-0.00) (-0.60) (0.51)Constant 1.32*** 1.96*** 1.82** -1.229 3.56*** 3.577*** (4.34) (4.51) (2.13) (-0.74) (3.92) (3.95)Year dummies Included Included Included Included Included IncludedAdj. R2/Pseudo R2 0.1912 0.1023 0.2408 0.3362 0.1249 0.3077F-statistics 4.15*** 3.47*** 8.45*** 5.93*** 1.76** 7.78***Number of banks 74 130 59 74 130 59No. of pooled obs. 305 672 345 305 672 345

37

38

Table 7: Determinants of the Board Structure for Small, Medium and Large Banks (contd.)

Pred DUAL (eq.3)

Variables sign Small Medium Large

SCOPE + .716* .578** -.011* (1.80) (2.45) (-1.73)EINDEX - -.118 -.163 -.372 (-0.23) (-0.52) (-0.88)MONCOSTS + .469 -.034 -.946*** (1.52) (-0.16) (-3.48)CEOPOWER + 1.23*** .71** 1.061** (3.00) (2.42) (2.26)CEOAGE + -3.01 -2.07 -.53 (-0.81) (-0.92) (-0.12)CEOWN OUTSIDEOWN - -.044* -.016 -.076*** (-1.67) (-0.82) (-2.84)NBLOCKOWN - -.03 .016 -.033* (-1.01) (0.51) (-1.88)BS t-1 INDIR t-1 - 1.23* -.5 1.23 (1.74) (-0.78) (0.78)CAPITAL - .383** -.3*** -.37*** (2.54) (-2.80) (-3.47)MERGERt-1 - -1.14** -.007* -1.2** (-2.42) (-1.82) (-2.35)DSOX - .234 -.57 -.379 (0.30) (-1.63) (-0.75)Constant 4.88 13.69 2.45 (0.33) (1.45) (0.12)Year dummies Included Included IncludedAdj. R2/Pseudo R2 0.3128 0.1214 0.2701F-statistics 3.50*** 1.55* 3.84***Number of banks 74 130 59No. of pooled obs. 305 672 345

Table 8: Univariate Results of the Changes in Bank Board Structure between Pre- and Post-SOX Periods.

This table presents changes in bank board structure variables between pre- and post-SOX periods. The pre- and post-SOX periods are represented by 2001 and 2004 respectively. Paired t-statistics show the significance of the difference in means between two matched pair samples, calculated following Selvanathan et al. (2004, pp. 456-461) whereas the Wilcoxon signed rank sum test-statistics show the significance of the difference in medians between two matched pair samples, calculated following Selvanathan et al. (2004, pp. 546-551). The size groups are formed by ranking the banks into quartiles based on their total assets each year. We label the first quartile banks “small”, quartiles second and third “medium” and the fourth quartile “large”Superscripts *, **, *** represent statistical significance at 10%, 5%, and 1% level respectively.

Pre-SOX

(2001) Post-SOX

(2004)

Difference (Post-SOX less Pre-

SOX) Paired t-statistics

Wilcoxon signed rank sum test -statistics

N 207 207 Board size (number) 12.93 12.37 -0.56 -3.326*** -2.990***Executive directors (%) 16.01 14.56 -1.45 -3.094*** -3.436***Independent directors (%) 63.40 66.59 3.19 4.665*** 4.263***CEO duality (dummy) 0.57 0.55 -0.02 -0.666 -0.581Total assets ($ in mil.) 22,683.17 33,495.07 10,811.90 2.794*** 12.121***

39

Table 9: Determinants of Bank Board Structure for Pre- and Post-SOX This table presents the results of the pooled ordinary least squares (OLS) estimates of equations (1) and (2) and LOGIT estimates of equation (3). Interactions with DSOX terms are included in the respective determinants equations. Standard errors in all estimations are clustered by banks. BS is the number of directors on the board. INDIR is the independent directors as a percentage of board size. DUAL is the dummy variable which equals 1 if the CEO also chairs the board. SCOPE is the PCA factor covering TA, AGE and DIVER. MONCOST is the PCA factor for CV and RISK. CEOPOWER is the PCA factor for lag of ROA, CEOT, and CEOWN. CEOAGE is the age of the CEO in years. OUTSIDEOWN is the percentage of shares owned by the directors and top executives excluding CEOWN. NBLOCKOWN is the percentage of shares owned by non-affiliated persons/institutions with 5% or more of the bank’s equity. CAPITAL is the bank equity as percentage of total assets. MERGER is the dummy variable which equals 1 if the bank has any M&A in the period. DSOX is the dummy variable which equals 1 if the period is either 2003 or 2004. . Figures in parentheses show the robust t-statistics based on standard errors clustered by both bank and year. Figures in brackets present the p-values of the respective F-statistics from the Wald (1943) tests. Superscripts *, **, *** indicate statistical significance at 10%, 5%, and 1% levels, respectively.

Variables Pred. sign BS (1)

Pred. sign INDIR (2)

Pred. sign DUAL (3)

Panel A: Explanatory variables: SCOPE + 0.10***(7.12) + 0.027**(2.08) + 0.403***(3.62)EINDEX + 0.03*(1.25) + -0.0167(-0.72) - -0.238(-1.1)MONCOSTS - 0.014(0.94) - -0.0109(-0.57) + -0.122(-0.75)CEOPOWER - -0.0118(-0.67) + 0.513***(2.81)CEOAGE - 0.1099(0.9) + 2.699**(2.17)CEOWN - -0.0047**(-2.12) - OUTSIDEOWN + 0.0087***(3.97) + -0.009***(-4.27) - -0.04***(-2.91)NBLOCKOWN + 0.00048(0.2) - -0.019(-1.21)BS t-1 + 0.138***(2.67) INDIR t-1 + 0.175***(2.93) - 0.1266(0.28)CAPITAL + 0.0067(0.61) + -0.0017(-0.22) - -0.156**(-2.14)MERGERt-1 + 0.1185***(4.91) + 0.0178(0.8) - -0.452*(-1.97)DSOX +/- -.194***(-3.47) + 0.633(1.59) +/- -0.457**(-2.12)SCOPE*DSOX 0.000117(0.01) -0.0004(-0.05) 0.125(1.02)EINDEX*DSOX 0.027*(1.7) 0.0212(1.32) 0.559***(3.06)MONCOST*DSOX 0.0234(1.28) 0.0039(0.2) 0.2105(1.09)CEOPOWER*DSOX -0.0049(-0.55) 0.249*(1.69)CEOAGE*DSOX -0.1624*(-1.67) 0.1853(0.17)CEOWN*DSOX 0.00139(0.73) OUTSIDEOWN*DSOX 0.0028(1.62) 0.00058(0.39) 0.0143(0.99)NBLOCKOWN*DSOX 0.000328(0.28) 0.0027(0.28)CONSTANT 1.489***(5.45) 3.48***(6.6) -10.61*(-1.68)Year dummies Included Included IncludedPanel B: Model fits Adjusted R2/Pseudo R2 0.2298 0.1905 0.1699F-statistics (19|23|23, 211) 10.51***[0.00] 4.15***[0.00] 3.29***[0.00]Number of banks 212 212 212No. of pooled obs. 1322 1322 1322

40

Endogenously structured boards of directors in banks

Documents

Transcript of Endogenously structured boards of directors in banks