Artikel SMK (Bank)

Journal of Banking & Finance 34 (2010) 15901606
Contents lists available at ScienceDirect
Journal of Banking & Finance

journal homepage: www.elsevier.com/locate/jbf
Endogenously structured boards of directors in banks

Shams Pathan a,*, Michael Skully b
a b
UQ Business School, The University of Queensland, Brisbane, Queensland 4072, Australia Department of Accounting and Finance, Monash University, Melbourne, Victoria 3145, Australia
a r t i c l e
i n f o
a b s t r a c t
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 benets 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 ofcers 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 specications. Furthermore, the studys ndings have important policy implications for bank regulators and investors. 2010 Elsevier B.V. All rights reserved.
Article history: Received 4 September 2009 Accepted 7 March 2010 Available online 10 March 2010 JEL classication: G21 G28 G30 G32 G34 L22 K22 Keywords: Board of directors Independent directors CEO CEO duality Endogenous Bank holding companies
1. Introduction A wide range of accounting, nance 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 SarbanesOxley 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 Alchians (1950) economic theory of Darwinism, it is important to understand why some rms 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 benets
* Corresponding author. Tel.: +61 7 3346 8075; fax: +61 7 3346 8166. E-mail addresses: s.pathan@business.uq.edu.au (S. Pathan), michael.skully@ buseco.monash.edu.au (M. Skully). 0378-4266/$ - see front matter 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.jbankn.2010.03.006
associated with boards monitoring and advising functions (Hermalin and Weisbach, 1998; Raheja, 2005; Adams and Ferreira, 2007; Harris and Raviv, 2008). Based on these theoretical works, among others, Baker and Gompers (2003), Boone et al. (2007), Coles et al. (2008), Linck et al. (2008), and Lehn et al. (2009) empirically nd evidence in support of the endogenous formation of boards of non-nancial rms. While the same theoretical underpinnings relating to board structure are valid to both banks and non-bank rms, the existing empirical studies exclude banks from their sample, and several factors (such as regulation, high leverage) could limit generalizing non-nancial board structure ndings to banks. This study aims to ll this knowledge gap by investigating whether the costs and benets 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 nancial 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
1
CEO duality refers to a situation in which the CEO is also the board chair.
S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606
1591
as a governance mechanism than its non-bank counterparts because banks have become larger, complex and more diversied, following the deregulation with the RiegleNeal Interstate Banking and Branching Efciency Act of 1994, as well as the Gramm LeachBliley 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 subject to stringent regulatory scrutiny, compared to non-bank board directors. This regulation is commonly justied 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 (Demirg-Kunt and Detragiache, 2002). This constant regulatory monitoring could limit bank managers self-serving behavior (such as perks). Hence, bank managers, including CEOs, cannot inuence the director selection process. As a result, in contrast to non-bank studies, CEO power (i.e., CEOs ability to inuence 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 benets 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 signicantly for regulated banks due to the enactment of the SarbanesOxley 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 nds some evidence in favor of endogenously chosen boards of directors. This supports the argument that banks structure their boards consistently with the costs and benets associated with boards monitoring and advising functions. More specically, the results show that: (i) larger and more diversied 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 benets 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 inuence and the constraints on managerial inuence 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 signicant 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 rm evidence. However, the percentage of independent directors increases substantially, especially during the post-SOX period. This study contributes to the existing literature on board structure determinants in several important ways. This is the rst study to demonstrate that even in a regulated industry like banking, the costs and benets of monitoring and advising functions of boards could explain their structure. This paper complements and extends Adams and Mehrans (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 inuence 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 conrm 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 rst study to provide some evidence that bank board structure has signicantly changed, following SOX and the associated changes mandated by the stock exchanges. Such ndings 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.
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 rms. The duciary2 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 Macey and OHara, 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 ofcers for their decisive roles (see Shepherd, 1992 for details). For example, Section 1821(k) of FIRREA 1989 stated that directors and ofcers 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 Ofce of the Comptroller of the Currency (OCC), 12 involved senior bank ofcers, including CEO and directors.3 Thus, bank directors and managers have a legal duty to recognize their obligation to the
2 Macey and Miller (1993, pp. 401407) dene duciary duties as . . . the mechanism invented by the legal system for lling in the unspecied terms of shareholders contingent [contracts]. In addition, see Macey and OHara (2003, pp. 9395) for a good discussions of bank directors duciary 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.
1592
debt claimants, especially when a bank is in a weakened nancial condition. Fanto (2006, p. 6) uses the metaphor that bank regulators stay with bank directors and ofcers from the cradle to the grave to illustrate how intensively bank regulators monitor and discipline bank management. In this regard, Pathan (2009) shows that banks in which CEOs have more power to inuence board decision take less risk. It should also be noted that certain regulations at the bank level, as opposed to the BHCs 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 525 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 banks head ofce, 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 specic 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 prots are in excess of $50 million), while two-thirds of these directors should be nonexecutives, i.e., outsiders.4 During this studys 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 inuence the banks 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 specic 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 benets of the board monitoring and advising roles, along with other rm 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., 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 selfserving 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 specic 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, 3.13.4, 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 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;
Source: www.banking.state.ny.us. 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 4
Linck et al., 2008). In comparison to small rms, large and diversied rms require additional board members to support their complex and diversied activities (Agrawal and Knoeber, 1996), and also to monitor management performance (Coles et al., 2008; Lehn et al., 2009). Large rms could also require more directors to serve on their boards sub-committees, handling the nomination of board members, compensation, and auditing (Boone et al., 2007). Similarly, the information requirements of larger and more complex rms generally result in the need for larger boards. In this regard, prior studies have established a positive relationship between board size and the rms 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 rms 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 rms 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 rms, since they bring valuable expertise and potential networks that could be benecial to the rm. They further contend that even though monitoring costs6 increase with a rms scope of operations, the benets 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 rm 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 diversied 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 rst Hypothesis H1 related to the scope of operations is as follows: Hypothesis H1. Bank board size and the percentage of independent directors are positively related to the banks 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 benets of monitoring managers private benets,7 as well as the monitoring costs to directors (Raheja, 2005; Adams and Ferreira, 2007; Harris and Raviv, 2008). With regard to private benets, the benet gained from monitoring managers increases if managers have opportunity to extract more private benets from their rms (Boone et al., 2007; Chi and Lee, 2010). Generally, for managers, such opportunities arise when rms have free cash ows, 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 difcult, if not impossible, to remove all the directors at the
6 The term monitoring costs is used to describe the costs related to information acquisition and processing in transforming directors expertise to the specic rms for which they serve as directors (Linck et al., 2008, p. 311). 7 The term private benets is dened 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.
1593
same time. In the presence of such opportunity for greater private benets 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 rms with high information asymmetry (Fama and Jensen, 1983). Prior studies suggest that rms with greater monitoring costs should rely less on outside directors (Fama and Jensen, 1983; Lehn et al., 2009). It is costly to transfer rm-specic information to outsiders because they are less informed about the rms 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 rm-specic 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, rms with high information asymmetry could benet from smaller board size and a greater representation of inside directors because of high monitoring costs. Generally, information asymmetry is high for rms with high stock return volatility (Fama and Jensen, 1983), high growth potential (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 benets are high and the cost of monitoring is low. Accordingly, Boone et al. (2007) nd a statistically signicant positive (negative) relation between monitoring private benets (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 benets on board size, but instead nd a statistically signicant positive relation between monitoring private benets 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 benets 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 CEOs (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 CEOs perceived ability (relative to a replacement) to inuence board decisions, as can be proxied by rm performance, CEO tenure or CEO ownership. Likewise, CEOs (negotiation) power can be constrained by the presence of shareholdings of non-afliated block-holders or outside directors on the board (Boone et al., 2007). Board independence could also be inuenced 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 (Vancil, 1987; Brickley et al., 1997). Using the horse races argument, the rm 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 (Vancil, 1987). This succession process indicates that board inde-
pendence decreases as the CEO approaches retirement (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 inuence 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 $1000 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 rm 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 reect the rms ownership incentives structure (Rediker and Seth, 1995; Raheja, 2005; Linck et al., 2008). This ownership incentives structure explains that variations in the rms ownership structure can also be important in aligning both managers and shareholders interests (McConnell and Servaes, 1990; Tong, 2008). 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 rms prot. These shareholdings reduce project verication costs, and potentially communication and coordination costs as well. Rahejas (2005, Proposition 6, p. 296) model, therefore, species that an optimal board is larger and has a majority of independent directors when verication costs are low. However, Linck et al. (2008) nd 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) nd that the proportion of outside directors is negatively related to both the non-afliated block-holders shareholdings and the managerial shareholdings. However, consistent with Rahejas (2005) proposition for banks, Whidbee (1997) shows that the proportion of outside directors is negatively related to CEO ownership, while positively related to non-afliated 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.
1594
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 ndings can nevertheless help in formulating some specic expectations regarding CEO duality (e.g., Hermalin and Weisbach, 1998; Linck et al., 2008). As an insider, the CEO possesses rm-specic knowledge that is important for large, complex and diversied banks. Likewise, banks with high monitoring costs, i.e., with high information asymmetry, could benet from CEO duality. It could also be argued that in the presence of opportunities for insiders to extract private benets, 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 banks scope of operations. Hypothesis H5B. CEO duality is negatively related to management private benets and positively related to directors monitoring costs. Consistent with Hermalin and Weisbachs (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) nd that the probability of CEO duality increases with CEO power and closeness of 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 specied 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.
tations 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 Financials 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 nal 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 scal year. INDIR is the number of independent directors, as a percentage of the total number of board directors. Following prior studies (e.g., Hermalin and Weisbach, 2003; Adams and Mehran, 2009), this study denes independent directors as those whose only business relationship with the bank is their directorship. When evaluating directors 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 diversication index. Bank size (TA) is the banks total assets, and bank age (AGE) is the number of years since a bank was rst listed on DATASTREAM. Following Stiroh and Rumble (2006, p. 127), bank revenue diversication 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 diversication for non-bank rms. However, for banks, Stiroh and Rumbles (2006) revenue diversication approach seemed more appropriate because it captures the complexity and the level of diversication of banks through their income sources. DIVER measures the degree of diversication in a BHCs 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 diversication for non-bank rms. However, for banks, Stiroh and Rumbles (2006) revenue diversication approach seemed more appropriate because it captures the complexity and the level of diversication of banks through their income sources. A shareholders restrictive rights index (EINDEX), an approximation of Bebchuk, Cohen and Ferrells (2009) entrenchment index,10 is used as a proxy for managers opportunities for private benets. 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
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. Denitions 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.
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 classication (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 SECs EDGAR lings. 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 scal year. The data collection procedure is then adjusted to account for when the proxies disclose some governance information for the previous scal year (e.g., the percentage of CEO shareholding) and others for the following scal year (e.g., the number of directors). The nancial 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 compu-
1595
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 costs are bank charter value (CV) and bank risk (RISK). CV is calculated following Keeleys 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 scal 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 (Jonghe and Vennet, 2008) and hence alters the need for monitoring by independent directors. 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 banks 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 banks total outstanding shares owned by the bank CEO (CEOWN). CEO succession planning is proxied by the CEOs 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-afliated block ownership (NBLOCKOWN). OUTSIDEOWN is calculated as the number of shares held by the bank ofcers 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-afliated block holders who own 5% or more shares as a percentage of total number of outstanding shares.12 Non-afliated block holders exclude inside directors and any trust company holding stocks on behalf of the banks employee stock ownership plan because the bank managers may have inuence over those trust companies through the contractual nature of their relationships. Several additional variables are included as control variables to reduce biases in the coefcient estimates due to omitted variables. These additional bank-specic 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 banks 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, 1998). 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 reect the addition
11 Boone et al. (2007) use the percentage of total outstanding shares owned by outside directors, rather than by ofcers 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 signicant at 1% level for a random sample of 40 banks. 12 SEC requires banks to disclose those shareholders owning at least ve percent of the rms total outstanding shares in the DEF 14A proxy statement. The shareholdings of others with less than ve 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).
of directors from an acquired banks board into the merged/acquirer banks 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 coefcient on DSOX also complements the univariate test results in Section 6, showing if bank board variables changes were statistically signicant 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, Eqs. (1)(3), are specied 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.
8 > a b1 lnTA i;t b2 lnAGE i;t b3 DIVER i;t > > > > > d1 EINDEX d2 CV d3 RISK > < i;t i;t i;t > c1 CEOWN i;t c2 OUTSIDEOWN i;t f1 INDIR i;t1 > > > > > f CAPITAL f MERGER > 2 3 : i;t i;t1 u DSOXi;t ei;t 8 > a b1 lnTA i;t b2 lnAGE i;t b3 DIVER i;t > > > > > d EINDEX d CV d RISK > 1 2 3 > i;t i;t > i;t > > > > /1 ROA i;t1 /2 CEOT i;t /3 CEOWN i;t > <

lnBSi;t
INDIRi;t
DUALi;t
> /4 lnCEOAGE i;t k1 OUTSIDEOWN i;t > > > > > k NBLOCKOWN f lnBS > f2 lnCAPITAL i;t > 2 1 i;t > i;t1 > > > > f3 MERGER > u DSOXi;t ei;t : i;t1 8 > a b1 lnTA i;t b2 lnAGE i;t b3 DIVER i;t > > > > > d EINDEX d CV d RISK > 1 2 3 > i;t i;t > i;t > > > > /1 ROA > i;t1 /2 CEOT i;t /3 CEOWN i;t <
> > > > > > > > > > > > > > :
/4 lnCEOAGE i;t k1 OUTSIDEOWN i;t

k2 NBLOCKOWN i;t f1 INDIR i;t1

f2 lnCAPITAL i;t f3 MERGERi;t1 u DSOXi;t ei;t

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. b, d, /, k, c, f, u, and W are the parameters to be estimated. e is the idiosyncratic error term. The denitions of the variables and the relevant hypothesized (or expected) signs in regression Eqs. (1)(3) are as already discussed in Sub-Section 4.2 and summarized in Table 1. 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 Eqs. (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 Eq. (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 Whites (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
1596
Table 1 Summary of denitions and predicted signs of variables for Eqs. (1)(3). Variables Panel A: Dependent variables BS INDIR DUAL Panel B: Explanatory variables TA AGE DIVER Denitions Equation The number of directors in the BHCs board The percentage of independent directors in the BHCs board A dummy variable that equals 1 if the CEO also chairs the BHCs board, otherwise zero The total assets of the BHC as at the end of each scal year The number of years since the BHC was listed in the DATASTREAM database Stiroh and Rumbles (2006) revenue diversication 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) The sum of two dummy variables: staggered boards, and poison pills. The dummy for staggered boards equals 1 if the BHCs board is classied, otherwise zero. The dummy for poison pills equals one if the BHCs board has the poison pill provision, otherwise zero Keeleys 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 The standard deviation of daily BHCs stock returns in a year The return on average total assets which is calculated as net income after tax as a percentage of the BHCs average total assets. Average total assets is simply the average of beginning and end of year BHCs total assets The number of years the BHC CEO has served in this position The BHC CEOs age in years The percentage of total outstanding shares owned by the BHC CEO The percentage of total outstanding shares owned by the BHC ofcers and directors excluding those of the CEO The percentage of total outstanding shares owned by non-afliated blockholders who hold at least 5% of outstanding shares The BHCs total equity as a percentage of total assets 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 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 b1 (+) b2 (+) b3 (+) b1 (+) b2 (+) b3 (+) BS (1) INDIR (2) DUAL (3)
EINDEX
d1 (+)
d1 (+)
d1 ()
CV
d2 ()
d2 ()
d2 (+)
RISK ROA
d3 ()
d3 () /1 ()
d3 (+) /1 (+)
CEOT CEOAGE CEOWN OUTSIDEOWN NBLOCKOWN Panel C: Other control variables CAPITAL MERGER DSOX
c1 () c2 (+)
/2() /3() /4 () k1 (+) k2 (+)
/2 (+) /3 (+) /4(+) k1 () k2 ()
f2 (+) f3 (+) u (+/)
f2 (+) f3 (+) u (+/)
f2 () f3 () u (+/)
This table summarizes the denitions and predicted signs of the variables in regression Eqs. (1)(3) related to the testing of Hypotheses H11, H2, H3, H4, H5A and H5B. Column 1 shows the list of variables and their denitions are in column 2. Columns 36 show coefcients along with their predicted signs in regression Eqs. (1)(3) for the determinants of respective BHC board size (BS), percentage of independent directors (INDIR), and CEO duality (DUAL).
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 specied as endogenizing board structure variables. Particularly, Eqs. (1)(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 of independent directors, INDIR, of 64.52 (66.67%) is similar to the 66.52% reported by Cornett et al. (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. The mean (median) age of the CEO, CEOAGE, is 56.09 (56.00) with consistent with that reported by Cornett et al. (2009). The mean (median) CEOWN is 4.41% (1.30%). In Panel C of Table 3, 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-specic variables (in Panels C and D of Table 2) are not described further. Table 3 presents the Pearson product-moment correlation coefcients among the governance and bank-specic 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, Fig. 1 below shows the time trends of bank board structure (i.e., board size, percentage of board independence, and CEO duality) from 1997 to
S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606 Table 2 Descriptive statistics. Variables Mean SD 4.54 8.73 15.72 0.44 0.47 0.72 0.49 7.74 7.32 8.8 9.96 8.54 105.78 0.07 1.9 0.09 9.01 0.32 1.2 0.51 Min. 5 37.5 10 0 0 0 0 0 34 0 0.19 0 0.16 0.94 3 0.06 1 0 0.65 6.24 1st Quartile 10 80 55.56 0 0 1 0 3 51 0.55 4.04 0 1.02 1.05 7.99 0.3 6 0 1.65 1 Median 12 86.96 66.67 1 0 1 1 7 56 1.3 7.24 0 2.07 1.09 9.09 0.36 12 0 2.02 1.22 2nd Quartile 15 90.91 75 1 1 2 1 14 60 3.46 13.36 5.7 7.66 1.13 10.16 0.43 22 0 2.53 1.47 Max. 31 100 96.55 1 1 2 1 46 85 65.19 83.32 98.5 1484.10 1.64 21.59 0.5 31 1 17.32 5.59 Skew. 0.96 1.49 0.58 1.11 0.7 0.12 0.31 1.18 0.4 3.72 2.7 4.9 8.23 1.82 1.34 0.53 0.46 2.45 4.81 1.01
1597
Kurt. 3.83 5.94 3.1 2.22 1.49 1.91 1.1 4.52 3.84 18.72 14.46 38.07 81.31 10.54 7.93 2.71 1.9 6.99 42.75 41.62
Panel A: Board structure variables BS (No.) 12.92 OUTDIR (%) 84.62 INDIR (%) 64.52 STAGG 0.74 POISON 0.34 EINDEX 1.08 Panel B: CEO characteristics DUAL 0.58 CEOT (Years) 8.85 CEOAGE (Years) 56.09 CEOWN (%) 4.41 Panel C: Ownership structure OUTSIDEOWN (%) 10.25 NBLOCKOWN (%) 3.67 Panel D: Bank-specic variables TA (in bil.) 23.66 CV 1.1 CAPITAL (%) 9.26 DIVER 0.36 AGE (Years) 13.73 MERGER 0.11 RISK (%) 2.26 ROA (%) 1.24
This table presents the distribution of variables by showing mean, standard deviation (SD), minimum (Min.), rst quartile (1st Quartile), median (Median), second quartile (2nd Quartile), skewness (Skew.), and kurtosis (Kurt.). See Table 1 for variable denitions.
Table 3 Correlation matrix. Variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 BS INDIR EINDEX DUALCEO CEOT CEOAGE CEOWN OUTSIDEOWN NBLOCKOWN LNTA CV CAPITAL DIVER AGE MERGER RISK ROA 1 1.00 2 0.16 1.00 3 0.01 0.05 1.00 4 0.08 0.14 0.04 1.00 5 0.00 0.05 0.18 0.25 1.00 6 0.01 0.01 0.17 0.22 0.42 1.00 7 0.15 0.12 0.14 0.11 0.18 0.03 1.00 8 0.08 0.36 0.25 0.19 0.08 0.02 0.06 1.00 9 0.04 0.04 0.05 0.03 0.06 0.04 0.04 0.10 1.00 10 0.38 0.33 0.01 0.24 0.04 0.01 0.15 0.34 0.07 1.00 11 0.02 0.05 0.03 0.09 0.03 0.01 0.13 0.18 0.01 0.31 1.00 12 0.02 0.01 0.07 0.07 0.04 0.08 0.02 0.13 0.01 0.07 0.17 1.00 13 0.28 0.17 0.02 0.20 0.01 0.05 0.00 0.19 0.02 0.51 0.08 0.07 1.00 14 0.18 0.32 0.05 0.22 0.05 0.08 0.10 0.37 0.09 0.68 0.29 0.08 0.39 1.00 15 0.05 0.00 0.02 0.07 0.05 0.02 0.07 0.02 0.04 0.12 0.02 0.08 0.01 0.00 1.00 16 0.1 0.18 0.12 0.04 0.02 0.04 0.09 0.08 0.03 0.24 0.13 0.01 0.18 0.17 0.03 1.00 17 0.05 0.02 0.03 0.05 0.00 0.05 0.00 0.14 0.04 0.20 0.59 0.40 0.06 0.23 0.04 0.19 1.00
This table shows Pearson pair-wise correlation matrix. Bold texts indicate statistically signicant at 1% level or better. See Table 1 for variables denitions.
2004. The three size groupings are developed by ranking the banks into quartiles based on their total assets in each year. Then, the rst quartile banks are grouped as small, the second and third quartiles as medium, and the fourth quartile large. 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 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 at 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 Eqs. (1)(3). The relevant diagnostic tests in Panel C of Table 4 are based on pooledOLS without any robust adjustment for residuals. The average variance ination factors (AVIF) across all the columns indicate that
1598
17.00 16.00 15.00 14.00 13.00 12.00 11.00 10.00 1997 1998
Panel A: Board Size

Small Medium Large
75.00 70.00 65.00 60.00 55.00 50.00
Panel B: % Independent Directors

Small Medium Large
1999
2000
2001
2002
2003
2004
1997
1998
1999
2000
2001
2002
2003
2004
Panel C: % of Banks with CEO is also Board Chair

0.90 0.85 0.80 0.75 0.70 0.65 0.60 0.55 0.50 0.45 0.40 1997 1998 1999 2000 2001 2002 2003 2004 Small Medium Large
Fig. 1. Bank board structure trends: 19972004. The sample includes 212 BHCs over 19972004, i.e. a total bank observations of 1534. The size groups are formed by ranking the banks into quartiles based on their total assets each year. We label the rst 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.
the multicollinearity among the regressors should not be a concern in estimating the regression equations.13 The White (1980) alternative test for heteroskedasticity (P1) shows statistically signicant LM-statistics for each regression, which conrms the presence of heteroskedasticity with normal OLS estimates. Likewise, the pooled-OLS estimates appear to suffer from rst-order serial correlation, as indicated by statistically signicant F-statistics across all regressions with Wooldridges (2006) test for rst-order serial correlation (P2). The presence of rst-order serial correlation in the panel data also indicates the presence of an unobserved rm-xed effect (Wooldridge, 2002, p. 176). These justify the pooled-OLS estimates of Eqs. (1) and (2) and LOGIT estimates of Eq. (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 xed- and time-effects in the estimates. In Panel B of Table 4, the regression equations are well tted with adjusted/pseudo Rsquared of 27.55%, 21.25%, and 16.74%, respectively, for BS, INDIR, and DUAL regressions, with statistically signicant F-statistics. With regard to Hypothesis 1, as mentioned earlier in Section 4.2, three variables are used to proxy for the banks 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 coefcients on TA and DIVER remain statistically signicant in the BS regression. The use of multiple proxies for an underlying variable in one regression equation could bias the coefcient toward zero, which is commonly known as attenuation bias (Lubotsky and Wittenberg, 2006). Similarly, the coefcients on TA and AGE are still positive and statistically signicant for the INDIR regression. More to the point, the Wald (1943) tests for the joint signicance of the coefcients 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 signicant F-statistics. Thus, Hypothesis 1 is well-supported and conrms that
bank board size and the percentage of independent directors are positively related to the banks scope of operations. With regard to the monitoring hypothesis, i.e., Hypothesis 2, in the BS regression, the coefcient on the EINDEX is positive and statistically signicant at the 5% level, and the coefcients on the CV and RISK are both negative but statistically signicant for CV at 10%. In the INDIR regression, while the coefcient on EINDEX is not statistically signicant, the coefcients on CV and RISK are both negative and statistically signicant at 10% or better.14 The Wald (1943) tests for the joint signicance of these two coefcients (CV and RISK) of the monitoring costs (in Panel D of Table 4) produce statistically signicant F-statistics for INDIR regression, but not for BS regression. Thus, the overall monitoring hypothesis (Hypothesis 2) is partially supported, as board size is positively related to the monitoring managements private benets, 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., Hypothesis 3), as expected, none of the negative coefcients on all of the three measures of CEO power (lag ROA, CEOT, and CEOWN) and a positive coefcient on the proxy for CEO succession planning (CEOAGE) is statistically significant in the INDIR regression. Similarly, the Wald (1943) test statistic for the joint signicance of these four variables (lag ROA, CEOT, CEOWN and CEOAGE) is not statistically signicant (F-statistic = 0.40, p-value = 0.81). Contrary to the prediction, the coefcients on the two proxies for constraints on CEO power, OUTSIDEOWN and NBLOCKOWN, are negative and statistically signicant for the former at the 1% level. The results of OUTSIDEOWN and NBLOCKOWN, however, are consistent with ndings by Linck et al. (2008) for non-bank rms 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
14 In an unreported regression where RISK is the only proxy for monitoring costs, along with other relevant variables, a statistically signicant negative coefcient on RISK is indicated.
According to Chatterjee et al. (2000), the guidelines for detecting multicollinearity are: (i) the largest VIF is greater than 10, and (ii) the mean VIF is larger than 1.
13
S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606 Table 4 Regression results for the determinants of the bank board structure. Variables Panel A: Explanatory variables TA AGE DIVER EINDEX CV RISK ROAt1 CEOT CEOWN CEOAGE OUTSIDEOWN NBLOCKOWN BSt1 INDIRt1 CAPITAL MERGERt1 DSOX Constant Panel B: Model ts Adjusted R2/pseudo R2 F-stats. (13|16|17, 1309) No. of pooled obs. Panel C: Regression diagnostics AVIF (max.) P1: LM-stats (v2 = 2) P2: F-stats (1, 211) Panel D: Wald test (F stat) for joint signicance of SCOPE, F (3, 1309|1305) MONCOST, F (2, 1309|1305) INCENTIVE, F (2, 1309) CEOPOWER & SUCCESSION, F (4, 1305) Constraints on CEO power, F (2, 1305) Pred. + + + + BS (1) 0.087***(6.74) -0.048(-1.61) 0.596***(3.36) 0.05**(2.09 0.367*(1.68) 0.005(0.50) Pred. + + + + + + + + + + INDIR (2) 0.021*(1.72) 0.033*(1.67) 0.115(0.62) 0.016(0.74) 0.350**(2.02) 0.034*(1.75) 0.008(0.29) 0.001(0.58) 0.001(0.72) 0.037(0.30) 0.009***(5.09) 0.0003(0.14) 0.123***(2.77) 0.0009(0.14) 0.007(0.37) 0.009*(1.78) 4.042***(8.2) 0.2125 19.81***[0.00] 1322 1.41 (2.50) 89.03*** [0.00] 29.75*** [0.00] 3.15** [0.02] 2.89* [0.06] N/A .40 [0.81] 13.03*** [0.00] Pred. + + + + + + + + + + +/ DUAL (3)
1599
0.004***(2.59) 0.009***(4.53)
0.225*(1.68) 0.162(0.63) 2.568*(1.68) 0.039(0.18) 1.892**(1.97) 0.0413(0.47) 0.127(0.70) 0.069***(3.10) 0.035**(1.75) 2.416**(2.23) 0.0339**(2.53) 0.016(1.16) 0.213(0.48) 0.144**(2.16) 0.487**(2.25) 0.301**(2.32) 14.19***(2.80) 0.1674 29.84***[0.00] 1322 1.39 (2.28) 85.71*** [0.00] 74.61*** [0.00] 6.90*** [0.00] 1.59* [0.07] N/A 6.86*** [0.00] 3.67** [0.03]
+ + +/
0.167***(3.11) 0.013*(1.72) 0.095***(2.96) 0.06***(3.62) 1.164***(2.97) 0.2755 55.57***[0.00] 1322 1.34 (2.25) 13.33*** [0.00] 47.94*** [0.00] 33.72***[0.00] 1.63 [0.20] 17.20*** [0.00] N/A N/A
This table presents the results of the pooled ordinary least squares (OLS) estimates of Eqs. (1) and (2) and LOGIT estimates of Eq. (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 scal year-end. AGE is the number of years since the BHC was listed in the DATASTREAM database. DIVER is the revenue diversication 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 banks 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-afliated persons/institutions with 5% or more of the banks 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 ination factor shows the degree of collinearity problem among the regressors. P1 is the White (1980) test for heteroskedasticity which provides the Lagrange Multiplier (LM) statistics based on alternative procedure explained in Wooldridge (2006, pp. 282283). P2 is the test for rst-order serial correlation which provides an F-statistic based on Wooldridge (2002, pp. 282283). Finally, the Wald (1943) test is used to assess the joint signicance of the respective estimates. Figures in parentheses show the robust tstatistics 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. * Statistical signicance at 10% levels. ** Statistical signicance at 5% levels. *** Statistical signicance at 1% levels.
has stronger incentives to monitor. Thus, as anticipated, Hypothesis 3 is not supported for banks. That is, bank board 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 (Hypothesis 4) in BS regression, as anticipated, the coefcient on CEOWN is negative and statistically signicant at the 1% level. Likewise, the coefcient on OUTSIDEOWN is positive and statistically signicant at the 1% level. The Wald (1943) test for the joint signicance of the two variables, CEOWN and OUTSIDEOWN (in Panel D of Table 4), shows a statistically signicant F-statistic of 17.20 (p-value <0.01). Thus, Hypothesis 4 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 Eq. (3), show that the coefcients on all three measures of scope of bank operations, TA, AGE and DIVER, are positive and statistically signicant for TA
and DIVER. The statistically signicant F-statistics with Wald (1943) test (in Panel D of Table 4) for the joint signicance of TA, AGE and DIVER also conrms Hypothesis H5A that the probability of CEO duality increases with banks scope of operation. However, the coefcient on EINDEX is negative, but not statistically signicant. The coefcient on CV is positive and statistically signicant at the 5% level, while the coefcient on RISK is not statistically signicant. Yet, for non-bank rms, Linck et al. (2008) nd no relation between CEO duality and directors monitoring costs. The Wald (1943) test indicates the joint signicance 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 ndings 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 benets. With regard to Hypothesis H5C, the coefcients on all the four measures of both CEO power and succession planning (lag ROA, CEOT, CEOWN and CEOAGE) are positive and statistically signicant at the 5% level or greater for CEOT, CEOWN and CEOAGE. As
1600
anticipated, the coefcients on OUTSIDEOWN and NBLOCKOWN, the proxy for constraints on CEO power, are both negative and statistically signicant at 5% for OUTSIDEOWN. The statistically significant Wald (1943) test statistics for the joint signicance 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 H5B 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. The results for control variables yield further insights into the forces shaping bank boards. The statistically signicant positive coefcient 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 signicant coefcients on lag INDIR in BS regression show that bank board size is positively related to the prior period percentage of independent directors. The statistically signicant coefcients 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 banks board are likely to be added to the acquirer banks board, in order to attract the target to accept the bid offer and also to share the board leadership
Table 5 3SLS regression results for the determinants of the bank board structure. Variables Panel A: Explanatory variables TA AGE DIVER EINDEX CV RISK ROAt1 CEOT CEOWN CEOAGE OUTSIDEOWN NBLOCKOWN BSt1 INDIRt1 CAPITAL MERGERt1 DSOX Constant Panel B: Model ts Adjusted R2/pseudo R2 v2-stats. (13|16|16, 1534) No. of banks No. of pooled obs. Panel C: Wald test (F-stat.) for joint signicance of SCOPE, F (4|3|3, 211) MONCOST, F (2, 211) INCENTIVE, F (2, 211) CEOPOWER & SUCCESSION, F (4, 211) Constraints on CEO power, F (2, 211) Pred. sign + + + + BS (Eq. (1)) 0.084***(11.13) 0.052(1.11) 0.609***(6.03) 0.052***(4.42) 0.319***(2.64) 0.008(1.13)
structure. Finally, the statistically signicant coefcients 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), Eqs. (1)(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 specied by Eq. (1). The ndings remain the same as
Pred. sign + + + + + + + + + +
INDIR (Eq. (2)) 0.014**(1.99) 0.035***(2.70) 0.161(1.56) 0.0187(.76) 0.303**(2.46) 0.0329***(5.07) 0.009(0.51) 0.001(0.95) 0.001(-1.24) 0.035(0.57) 0.009***(11.08) 0.0003(0.33) 0.202***(8.57) 0.0002(0.05) 0.008(0.33) 0.0112(0.68) 3.883***(13.63) 0.2059 403.48***[0.00] 212 1532 6.86***[0.00] 16.24***[0.00] N/A 3.04 [0.17] 62.36***[0.00]
Pred. sign + + + + + + + + + + +/
DUAL (Eq. (3)) 0.0430***(3.7) 0.033(1.47) 0.494***(3.17) 0.0006(0.03) 0.365*(1.71) 0.007(0.63) 0.005(0.15) 0.0128***(7.14) 0.006***(4.32) 0.471***(4.41) 0.007***(4.63) 0.003**(2.29) 0.047(1.06) 0.029***(4.04) 0.10**(2.51) 0.06**(2.14) 2.295***(4.42) 0.2037 337.35***[0.00] 212 1532 22.37***[0.00] 2.74*[0.07] N/A 49.72***[0.00] 12.45***[0.00]
0.004***(4.43) 0.01***(10.13)
+ + +/
0.26***(8.99) 0.012***(2.85) 0.096***(3.62) 0.06***(3.22) 0.741*(3.88) 0.2598 525.55***[0.00] 212 1532 90.12***[0.00] 4.25***[0.01] 63.78***[0.00] N/A N/A
This table presents the results of the three-stage least squares (3SLS) estimation of the system of simultaneous Eqs. (1)(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 scal year-end. AGE is the number of years since the BHC was listed in the DATASTREAM database. DIVER is the revenue diversication 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 banks 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-afliated persons/institutions with 5% or more of the banks 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 signicance 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. * Statistical signicance at 10% levels. ** Statistical signicance at 5% levels. *** Statistical signicance at 1% levels.
1601
with those reported in Table 4 with improve statistical signicance. Column 5 of Table 5 presents the ndings for the percentage of independent directors (INDIR). The interpretation of the results also remains the same as those in Table 4, except that the coefcient on DSOX is no longer statistically signicant. The ndings for CEO duality (specied by Eq. (3)) in Column 7 of Table 5 also do not show any signicant deviations from the results reported in Table 4. However, the negative coefcient on NBLOCKOWN is now statistically signicant. Thus, even with direct control for endogeneity using 3SLS, this study nds evidence that the costs and benets of the boards 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) Table 6 below presents Arellano and Bover (1995) and Blundell and Bond (1998) two-step system GMM estimates of Eqs ()()()(1) (3). In the system GMM, rst-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 tted with statistically insignicant test statistics for both second-order autocorrelation in second differences (P2) and Hansen J-statistics of over-identifying restrictions. The residuals in the rst difference should be serially correlated (P1) by way of construction but the residuals in the second difference should not be serially correlated (P2). Accordingly, in Panel B of Table 6, we could see statistically signicant P1 and statistically insignicant P2 for all equations. Likewise, the Hansen J-statistics of over-identifying restrictions tests the null of instrument validity and the statistically insignicant 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 Eq. (1) and 82 for both Eqs. (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 Eq. (1) for BS, the coefcient on EINDEX is no longer statistically signicant. The coefcients on both RISK and OUTSIDEOWN are not also statistically signicant in Eq. (2) for INDIR. Likewise, in the Eq. (3) for DUAL, the coefcient on CEOAGE is not statistically signicant while the positive coefcient on CV is now statistically significant. However, these differences do not discredit the overall
Table 6 Two-step system GMM regression results for the determinants of the bank board structure. Variables Panel A: Explanatory variables TA AGE DIVER EINDEX CV RISK ROAt1 CEOT CEOWN CEOAGE OUTSIDEOWN NBLOCKOWN BSt1 INDIRt1 CAPITAL MERGERt1 DSOX Constant Year dummies Panel B: Model ts F-statistics (16|20|20, 211) P1 P2 Hansen J-statistics (v2 = 156|60|60) No. of instruments No. of banks No. of pooled observations Pred. sign + + + + BS (1) 0.084***(4.39) 0.052(1.24) 0.48***(2.67) 0.025(0.75) 0.84**(2.01) 0.006(0.52) Pred. sign + + + + + + + + + + INDIR (2) 0.024*(1.84) 0.053*(1.86) 0.084(0.43) 0.011(0.42) 0.553*(1.70) 0.02(0.80) 0.001(0.03) 0.001(0.04) 0.0003(0.07) 0.168(0.60) 0.006(1.41) 0.003(0.81) 0.033*(1.86) 0.008(0.77) 0.002(0.12) 0.007*(1.72) 5.27***(3.90) Included 2.61 [0.00] 2.37**[0.018] 0.54 [0.586] 43.89 [0.941] 82 212 1322 Pred. sign DUAL (3) 0.069**(2.41) 0.001(0.04) 0.53*(1.72) 0.006(0.10) 0.191*(1.68) 0.013(0.54) 0.049(1.31) 0.0185***(2.61) 0.001*(1.67) 0.486(1.09) 0.009(1.26) 0.007(0.75) 0.059(0.37) 0.025*(1.79) 0.0457*(1.69) 0.153***(3.47) 1.8984(0.93) Included 3.99 [0.00] 3.15***[0.00] 1.03 [0.304] 58.57 [0.528] 82 212 1322
0.007*(1.89) 0.003*(1.72)
+ + + + + +
+ + +/
0.117*(1.69) 0.009(0.67) 0.068***(4.41) 0.07**(1.98) 2.28***(3.23) Included 8.72 [0.00] 3.40***[0.00] 0.22 [0.829] 168.20 [0.239] 174 212 1322
This table presents the results of the two-step system GMM (OLS) estimates of Eqs. (1)(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 scal year-end. AGE is the number of years since the BHC was listed in the DATASTREAM database. DIVER is the revenue diversication 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 banks 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-afliated persons/institutions with 5% or more of the banks 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, P1 and P2 are the test statistics for rst-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. * Statistical signicance at 10% levels. ** Statistical signicance at 5% levels. *** Statistical signicance at 1% levels.
1602
Table 7
Determinants of the board structure for small, medium and large banks.
Variables Pred sign SCOPE EINDEX MONCOSTS CEOPOWER CEOAGE CEOWN OUTSIDEOWN NBLOCKOWN BSt1 INDIRt1 CAPITAL MERGERt1 DSOX Constant Year dummies Adj. R2/pseudo R2 F-statistics Number of banks No. of pooled obs. Variables +/ + + +/ .28 (3.91) .006 (0.39) .084 (1.57) .07 (0.98) 1.32*** (4.34) Included 0.1912 4.15*** 74 305 Pred sign SCOPE EINDEX MONCOSTS CEOPOWER CEOAGE CEOWN OUTSIDEOWN NBLOCKOWN BSt1 INDIRt1 CAPITAL MERGERt1 DSOX Constant Year dummies Adj. R2/pseudo R2 F-statistics Number of banks No. of pooled obs. 1.23* (1.74) .383** (2.54) 1.14** (2.42) .234 (0.30) 4.88 (0.33) Included 0.3128 3.50*** 74 305 .5 (0.78) .3*** (2.80) .007* (1.82) .57 (1.63) 13.69 (1.45) Included 0.1214 1.55* 130 672 1.23 (0.78) .37*** (3.47) 1.2** (2.35) .379 (0.75) 2.45 (0.12) Included 0.2701 3.84*** 59 345 .044* (1.67) .03 (1.01) .016 (0.82) .016 (0.51) .076*** (2.84) .033* (1.88) + + + +
***
BS (Eq. (1)) Small .054* (1.69) .013 (1.76) .003 (0.11) Medium .088*** (2.70) .049* 1.68 .017 (0.62) Large .07* 1.74 .06* 1.74 .049 (0.98)
Pred sign + +
INDIR (Eq. (2)) Small 0.086* (1.87) 0.057 (1.16) 0.04** (1.96) 0.09 (1.58) 1.02 (0.82) Medium .058** (2.03) .026 (0.85) .013* (1.75) .004 (0.15) .13 (0.56) Large .037* (1.87) .024 (0.93) .014* (1.67) .023 (0.98) .12 (0.55)
+ +
.004 (1.77) .009*** (2.84)
.005 (1.84) .007* (1.89)
.011 (1.74) .011*** (4.07)
+ + +
0.014*** (3.53) 0.013*** (4.28) 0.51*** (3.46)
.007** (2.32) .003 (1.47) .049 (0.87)
.007*** (5.80) .004*** (3.27) .091 (1.16)
.073 (0.81) .015 (0.94) .13*** (4.00) .07 (1.37) 1.96*** (4.51) Included 0.1023 3.47*** 130 672
.123 (0.63) .011 (0.48) .15*** (3.02) .05 (1.30) 1.82** (2.13) Included 0.2408 8.45*** 59 345 DUAL (Eq. (3)) Small .716* (1.80) .118 (0.23) .469 (1.52) 1.23*** (3.00) 3.01 (0.81)
+ + +
0.002 (0.12) 0.021 (0.38) 0.0003 (0.00) 1.229 (0.74) Included 0.3362 5.93*** 74 305
.005 (0.49) .034 (1.08) .0189 (0.60) 3.56*** (3.92) Included 0.1249 1.76** 130 672
.003 (0.54) .01 (0.27) .017 (0.51) 3.577*** (3.95) Included 0.3077 7.78*** 59 345
Medium .578** (2.45) .163 (0.52) .034 (0.16) .71** (2.42) 2.07 (0.92)
Large .011* (1.73) .372 (0.88) .946*** (3.48) 1.061** (2.26) .53 (0.12)
This table presents the results of the pooled ordinary least squares (OLS) estimates of Eqs. (1) and (2) and LOGIT estimates of Eq. (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-afliated persons/institutions with 5% or more of the banks 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. * Statistical signicance at 10% levels. ** Statistical signicance at 5% levels. *** Statistical signicance at 1% levels.
1603
ndings regarding the determinants of board structure in banks given bank institutional arrangements as already reported in Section 5. 6.3. Results for other analysis Other robustness tests include Fama and MacBeth (1973) for cross-sectional dependence, iteratively re-weighted least squares (IRLS) for outliers, and Prais-Winsten (1954) for heteroskedasticity, cross-sectional dependence and rst-order serial dependence. The Fama and MacBeth (1973) estimates of regression Eqs. (1) (3), are robust to the contemporaneous cross-sectional dependence. In the rst step, for each year, a cross-sectional regression model is tted. In the second step, the nal coefcients are estimated as the average of the rst step coefcient estimates. The Fama and MacBeth (1973) estimates are comparable to the pooled-OLS estimates in Table 4, and hence not tabulated. The statistical signicances of some estimates have improved. For example, the positive coefcients on TA and AGE in INDIR regression Eq. (2) are now statistically signicant at 1% or better. Following Coles et al. (2008) and Linck et al. (2008), to resolve difculties with non-normality in residuals from outliers, regression Eqs. (1)(3) are re-estimated using IRLS procedure. IRLS is derived from Hubers (1964) M-estimators (for maximum-likelihood) family and is resistant to any outliers or inuential observations. Following Kmenta (1986, pp. 318320), regression Eqs. (1)(3) are also estimated with the non-conventional Prais-Winsten (1954) procedure, in which standard errors and the variancecovariance matrix are robust to heteroskedasticity, cross-sectional dependence and rst-order serial dependence. Generally, the results of these two procedures conrm the ndings displayed in Tables 4 and 5 with improved statistical signicance levels and for this reason are not in a table.15 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 banks scope of operations) could bias the coefcient 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 ndings do not deviate with the pooled-OLS estimation of regression Eqs. (1) and (2) and the LOGIT estimation of Eq. (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 Eqs. (1)(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 rst 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 Eqs. (1)(3) for these three distinct size groups. The ndings appear to remain the same for SCOPE,
15
MONCOSTS and CEOPOWER across these groups, in relation to bank board structure. However, there are a few signicant differences. For example, the positive coefcient on lag INDIR in BS determinants (i.e. Eq. (1)) is statistically signicant for small banks. Similarly, the negative coefcient on NBLOCKOWN in INDIR determinants (Eq. (2)) is statistically signicant 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 coefcient on OUTSIDEOWN in DUAL regression Eq. (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 ndings.
7. Impact of reforms with SOX Several studies have examined the impact of reforms with SOX on non-nancial board structure and rm value (e.g, Linck et al., 2008). Particularly, Linck et al. (2008) show that both board size and independence increased during the post-SOX period for nonnancial rms. They also demonstrate that the results for board structure determinants do not alter notably in the post-SOX. Similarly, for nancial service rms (banks, thrifts institutions, insurance and securities), Akhigbe and Martin (2006) nd a favorable valuation effect of SOX for those with more independent boards and audit committees, nancial 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 signicant 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 signicantly 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.16 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 signicant paired t-statistic of 3.326 (p-value <0.01) and Wilcoxon statistic of 2.990 (pvalue <0.01). For banks, this decrease in board size could reect an increase public awareness and investors preference for small boards because the bank size increase was statistically signicantly over the sample period. Due to less coordination and communication costs, smaller boards are more efcient than larger boards (Yermack, 1996). It is not surprising to see that the percentage of executive directors on the bank board decreased by 9%, which is statistically signicant at the 1% level (paired t-statistic = 3.094 and Wilcoxon statistic = 3.436). Similarly, the percentage of independent bank
16 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 rms board structure.
The results are available from the author upon request.
1604
Table 8 Univariate results of the changes in bank board structure between pre- and post-SOX periods. Pre-SOX (2001) N Board size (number) Executive directors (%) Independent directors (%) CEO duality (dummy) Total assets ($ in mil.) 207 12.93 16.01 63.40 0.57 22,683.17 Post-SOX (2004) 207 12.37 14.56 66.59 0.55 33,495.07 Difference (Post-SOX less Pre-SOX) 0.56 1.45 3.19 0.02 10,811.90 Paired t-statistics 3.326*** 3.094*** 4.665*** 0.666 2.794*** Wilcoxon signed rank sum test -statistics 2.990*** 3.436*** 4.263*** 0.581 12.121***
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 signicance of the difference in means between two matched pair samples whereas the Wilcoxon signed rank sum test-statistics show the signicance of the difference in medians between two matched pair samples. The size groups are formed by ranking the banks into quartiles based on their total assets each year. We label the rst quartile banks small, quartiles second and third medium and the fourth quartile large. *** Statistical signicance at 1% level.
Table 9 Determinants of bank board structure for pre- and post-SOX. Variables Panel A: Explanatory variables SCOPE EINDEX MONCOSTS CEOPOWER CEOAGE CEOWN OUTSIDEOWN NBLOCKOWN BSt1 INDIRt1 CAPITAL MERGERt1 DSOX SCOPE*DSOX EINDEX*DSOX MONCOST*DSOX CEOPOWER*DSOX CEOAGE*DSOX CEOWN*DSOX OUTSIDEOWN*DSOX NBLOCKOWN*DSOX CONSTANT Year dummies Panel B: Model ts Adjusted R2/pseudo R2 F-statistics (19|23|23, 211) Number of banks No. of pooled obs. Pred. sign + + BS (1) 0.10***(7.12) 0.03*(1.25) 0.014(0.94) Pred. sign + + + + + + + + INDIR (2) 0.027**(2.08) 0.0167(0.72) 0.0109(0.57) 0.0118(0.67) 0.1099(0.9) 0.009***(4.27) 0.00048(0.2) 0.138***(2.67) 0.0017(0.22) 0.0178(0.8) 0.633(1.59) 0.0004(0.05) 0.0212(1.32) 0.0039(0.2) 0.0049(0.55) 0.1624*(1.67) 0.00058(0.39) 0.000328(0.28) 3.48***(6.6) Included 0.1905 4.15***[0.00] 212 1322 Pred. sign + + + + +/ DUAL (3) 0.403***(3.62) 0.238(1.1) 0.122(0.75) 0.513***(2.81) 2.699**(2.17) 0.04***(2.91) 0.019(1.21) 0.1266(0.28) 0.156**(2.14) 0.452*(1.97) 0.457**(2.12) 0.125(1.02) 0.559***(3.06) 0.2105(1.09) 0.249*(1.69) 0.1853(0.17) 0.0143(0.99) 0.0027(0.28) 10.61*(1.68) Included 0.1699 3.29***[0.00] 212 1322
0.0047**(2.12) 0.0087***(3.97)
+ + + +/
0.175***(2.93) 0.0067(0.61) 0.1185***(4.91) .194***(3.47) 0.000117(0.01) 0.027*(1.7) 0.0234(1.28)
0.00139(0.73) 0.0028(1.62) 1.489 (5.45) Included 0.2298 10.51***[0.00] 212 1322

***
This table presents the results of the pooled ordinary least squares (OLS) estimates of Eqs. (1) and (2) and LOGIT estimates of Eq. (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-afliated persons/ institutions with 5% or more of the banks 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. * Statistical signicance at 10% levels. ** Statistical signicance at 5% levels. *** Statistical signicance at 1% levels.
directors (INDIR) increased by 5%, from 63.40% in 2001 to 66.59% in 2004 and is statistically signicant 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 signicant. In contrast, Linck et al. (2008) reported a statistically signicant decrease in the percentage of non-bank rms 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 Eqs. (1)(3), the post-SOX dummy, DSOX, interacts with each of the hypothesized determinants. The results show only three notable differences. The coefcients on both EINDEX and its interaction term with DSOX are positive and statistically signicant in the BS regression Eq. (1). The Wald (1943) test for the joint signicance of EINDEX and its interaction term with the DSOX also provide statistically signicant 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 coefcient on CEOAGE is insignicant, but its interaction terms with DSOX in the INDIR regression Eq. (2) is negative and statistically signicant. The Wald (1943) test for the joint
1605
signicance of CEOAGE and CEOAGE*DSOX provides statistically insignicant 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 Eq. (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.
Acknowledgements This paper is prepared from the rst empirical chapter of the rst authors doctoral thesis at Monash University. The authors wish to thank Barry Williams, Richard Heaney, Rene Adams, an anonymous referee, Ike Mathur (the Editor), Mark Flannery, Maureen OHara, Benjamin Hermalin, Robert Faff, John Kose, Mark Harris, Mohamad Ariff, J. Wickramanayake, Mamiza Haq, Peter Verhoeven, Janice How, Robert Bianchi, Michele Meoli, John Nowland, John Chen, Mohammad Hayat, Colleen Puttee, John Zhang, Alexander Akimov, George Tanewski, Petko Kalev, Noel Gaston, Gulesekaran Rajaguru, Hardjo Koerniadi, Mohamad Belkhir, and seminar/conference participants at Bond University, Deakin University, Queensland University of Technology, University of Western Sydney, Grifth University, 21st Australian Finance and Banking Conference, 2009 Asian Finance Association Conference, and 2009 AFAANZ Conference for their helpful comments. The rst author acknowledges the AFAANZ 2009 travel grant to support this research. The usual caveats apply.
8. Conclusion This paper investigates the trends and the endogenously chosen bank boards of directors, with specic reference to the costs and benets 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 benets of monitoring, negotiations with the bank CEO and incentives alignments, after controlling for other bank specic characteristics. Using a sample of 212 BHCs over the 19972004 period, or 1534 bank year observations, this study nds 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 ndings, the results demonstrate that larger and more diversied banks have larger boards, more independent directors and combined leadership structure. In the presence of managers opportunities to consume private benets, the study indicates that banks benet from larger boards, while banks benet from more independent directors when the costs of monitoring managers are low. In contrast to non-bank evidence, it is argued that bank managers including CEOs do not inuence 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 conrms 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, specically 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 signicantly in the post-SOX period. For example, the mean percentage of independent directors increased with statistical signicance 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 efciency 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 inefciency 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 ndings 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.
References
Adams, R.B., Ferreira, D., 2007. A theory of friendly board. The Journal of Finance 62, 217250. Adams, R.B., Mehran, H., 2009. Corporate performance, board structure and their determinants in the banking industry. Journal of Financial Intermediation, forthcoming. Agrawal, A., Knoeber, C.R., 1996. Firm performance and mechanisms to control agency problems between managers and shareholders. Journal of Financial and Quantitative Analysis 31, 377397. Akhigbe, A., Martin, A.D., 2006. Valuation impact of SarbanesOxley: evidence from disclosure and governance within the nancial services industry. Journal of Banking and Finance 30, 9891006. Alchian, A., 1950. Uncertainty, evolution and economic theory. Journal of Political Economy 58, 211221. Anderson, R.C., Fraser, D.R., 2000. Corporate control, bank risk taking, and the health of the banking industry. Journal of Banking and Finance 24, 13831398. Arellano, M., Bover, O., 1995. Another look at the instrumental variables estimation of error-components models. Journal of Econometrics 68, 2951. Baker, M., Gompers, P.A., 2003. The determinants of board structure at the initial public offering. Journal of Law and Economics 46, 569598. Bebchuk, L.A., Cohen, A., Ferrell, A., 2009. What matters in corporate governance? Review of Financial Studies 22, 783827. Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics 87, 115143. Boone, A.L., Field, C.L., Karpoff, J.M., Raheja, C.G., 2007. The determinants of corporate board size and composition: an empirical analysis. Journal of Financial Economics 85, 66101. Brickley, J.A., Coles, J.L., Jarrell, G., 1997. Leadership structure: separating the CEO and chairman of the board. Journal of Corporate Finance 3, 189220. Chatterjee, S., Hadi, A.S., Price, B., 2000. Regression Analysis by Example. John Wiley and Sons, New York. Chi, D.J., Lee, D.S., 2010. The conditional nature of the value of corporate governance. Journal of Banking and Finance 34, 350361. Coles, J.L., Daniel, N.D., Naveen, L., 2008. Boards: does one size t all? Journal of Financial Economics 87, 329356. Cornett, M.M., McNutt, J., Tehranian, H., 2009. Corporate governance and earnings management at large US bank holding companies. Journal of Corporate Finance 15, 412430. Demirg-Kunt, A., Detragiache, E., 2002. Does deposit insurance increase banking system stability? An empirical investigation. Journal of Monetary Economics 49, 13731406. Fama, E.F., Jensen, M.C., 1983. Agency problems and residual claims. Journal of Law and Economics 26, 327342. Fama, E.F., MacBeth, J., 1973. Risk, return and equilibrium: empirical tests. Journal of Political Economy 71, 607636. Fanto, J.A., 2006. Paternalistic Regulation of Public Company Management: Lessons from Bank Regulation. Brooklyn Law School Legal Studies Paper No. 49 (Viewed on August 29, 2009). <http://ssrn.com/abstract=873667>. Flannery, M.J., 1998. Using market information in prudential bank supervision: a review of the US empirical evidence. Journal of Money, Credit and Banking 30, 273305. Galai, D., Masulis, R.W., 1976. The option pricing model and the risk factor of stock. Journal of Financial Economics 3, 5381. Harris, M., Raviv, A., 2008. A theory of board control and size. Review of Financial Studies 21, 17971832. Hermalin, B.E., Weisbach, M.S., 1998. Endogenously chosen boards of directors and their monitoring of the CEO. The American Economic Review 88, 96118.
1606
S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606 Raheja, C.G., 2005. Determinants of board size and composition: a theory of corporate boards. Journal of Financial and Quantitative Analysis 40, 283306. Rediker, K.J., Seth, A., 1995. Boards of directors and substitution effects of alternative governance mechanisms. Strategic Management Journal 16, 85 99. Roodman, D., 2006. How to do xtabond2: An Introduction to Difference and System GMM in Stata. Center for Global Development Working Paper Series. Shepherd, J., 1992. The liability of ofcers and directors under the Financial Institutions Reform, Recovery and Enforcement Act of 1989. Michigan Law Review 90, 11191144. Stiroh, K.J., Rumble, A., 2006. The dark side of diversication: the case of US nancial holding companies. Journal of Banking and Finance 30, 21312161. Tong, Z., 2008. Deviations from optimal CEO ownership and rm value. Journal of Banking and Finance 32, 24622470. Vancil, R.F., 1987. Passing the Baton: Managing the Process of CEO Succession. Harvard Business School Press, Boston. Wald, A., 1943. Tests of statistical hypotheses concerning several parameters when the number of observations is large. Transactions of the American Mathematical Society 54, 426482. Whidbee, D.A., 1997. Board composition and control of shareholder voting rights in the banking industry. Financial Management 26, 2741. White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48, 817838. Wooldridge, J.M., 2002. Econometric Analysis of Cross-Section and Panel Data. MIT Press, Boston. Wooldridge, J.M., 2006. Introductory Econometrics: A Modern Approach, third ed. Thomson/South Western, Mason. Yermack, D., 1996. Higher market valuation of companies with a small board of directors. Journal of Financial Economics 40, 185212.
Hermalin, B.E., Weisbach, M.S., 2003. Boards of directors as an endogenously determined institution: a survey of the economic literature. FRBNY Economic Policy Review 9, 726. Huber, P.J., 1964. Robust estimation of a location parameter. The Annals of Mathematical Statistics 35, 73101. Jonghe, O.D., Vennet, R.V., 2008. Competition versus efciency: what drives franchise values in European banking? Journal of Banking and Finance 32, 18201835. Keeley, M.C., 1990. Deposit insurance, risk and market power in banking. The American Economic Review 80, 11831200. Kmenta, J., 1986. Elements of Econometrics, 2nd ed. Macmillan, New York. Lehn, K., Patro, S., Zhao, M., 2009. Determinants of the size and structure of corporate boards: 19352000. Financial Management 38, 747780. Linck, J.S., Netter, J.M., Yang, T., 2008. A large sample study on board changes and determinants of board structure. Journal of Financial Economics 87, 308328. Lubotsky, D., Wittenberg, M., 2006. Interpretation of regressions with multiple proxies. Review of Economics and Statistics 88, 549562. Macey, J.R., Miller, J.P., 1993. Corporate stakeholders: a contractual perspective. The University of Toronto Law Journal 43, 401424. Macey, J.R., OHara, M., 2003. The corporate governance of banks. FRBNY Economic Policy Review 9, 91107. Marcus, A.J., 1984. Deregulation and bank nancial policy. Journal of Banking and Finance 8, 557565. McConnell, J.J., Servaes, H., 1990. Additional evidence on equity ownership and corporate value. Journal of Financial Economics 27, 595612. Pathan, S., 2009. Strong boards, CEO power and bank risk-taking. Journal of Banking and Finance 33, 13401350. Petersen, M.A., 2009. Estimating standard errors in nance panel data sets: comparing approaches. Review of Financial Studies 22, 435480.

Artikel SMK (Bank)

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Artikel SMK (Bank)

Uploaded by

Copyright:

Available Formats

Journal of Banking & Finance 34 (2010) 15901606

Contents lists available at ScienceDirect

Journal of Banking & Finance

Endogenously structured boards of directors in banks

S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606

S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606

S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606

S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606

S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606

/4 lnCEOAGE i;t k1 OUTSIDEOWN i;t

k2 NBLOCKOWN i;t f1 INDIR i;t1

f2 lnCAPITAL i;t f3 MERGERi;t1 u DSOXi;t ei;t

S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606

/2() /3() /4 () k1 (+) k2 (+)

/2 (+) /3 (+) /4(+) k1 () k2 ()

f2 (+) f3 (+) u (+/)

f2 (+) f3 (+) u (+/)

S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606

Panel A: Board Size

75.00 70.00 65.00 60.00 55.00 50.00

Panel B: % Independent Directors

Panel C: % of Banks with CEO is also Board Chair

S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606

S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606

S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606

.004 (1.77) .009*** (2.84)

.005 (1.84) .007* (1.89)

.011 (1.74) .011*** (4.07)

0.014*** (3.53) 0.013*** (4.28) 0.51*** (3.46)

.007** (2.32) .003 (1.47) .049 (0.87)

.007*** (5.80) .004*** (3.27) .091 (1.16)

S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606

The results are available from the author upon request.

S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606

0.175***(2.93) 0.0067(0.61) 0.1185***(4.91) .194***(3.47) 0.000117(0.01) 0.027*(1.7) 0.0234(1.28)

0.00139(0.73) 0.0028(1.62) 1.489 (5.45) Included 0.2298 10.51***[0.00] 212 1322

S. Pathan, M. Skully / Journal of Banking & Finance 34 (2010) 15901606

You might also like

0.014* (3.53) 0.013* (4.28) 0.51*** (3.46)

.007* (5.80) .004* (3.27) .091 (1.16)

0.175(2.93) 0.0067(0.61) 0.1185(4.91) .194***(3.47) 0.000117(0.01) 0.027*(1.7) 0.0234(1.28)