CEO Duality: Determinants and Consequences. Ghosh, Karuna, & Moonˆ. January (Preliminary draft please do not cite) Abstract

Transcription

1 CEO Duality: Determinants and Consequences Ghosh, Karuna, & Moonˆ January 2011 (Preliminary draft please do not cite) Abstract We examine whether a firm s operating environment influences the likelihood that the CEO is also appointed the Chair of the board of directors, and whether CEO-Chairs use their added power to extract rents from their firms. Specifically, we find that when a firm has a greater advisory need, the firm is more likely to appoint its CEO as Chair. However, we only find weak evidence of a positive relation between a firm s need for managerial initiatives for innovation and its CEO being Chair. Finally, we show that managerial rent extraction is multidimensional by providing evidence that CEO-Chairs extract rents from their firms by receiving excess compensation and greater job security but not via managing earnings or negotiating for greater reliance on earnings- versus stock-based measures to determine their compensation. Our study provides evidence consistent with both the optimal contracting and managerial power hypotheses. Zicklin School of Business, Baruch College, City University of New York C.T. Bauer College of Business, University of Houston ˆ Yonsei School of Business, Yonsei University 1

2 1. Introduction The role of the CEO as the chairman of the board of directors (hereafter Chair) has been a controversial topic. For example, powerful institutional investors including California Pension Retirement System (CalPERS) and New York City Pension Fund have strongly advocated for the separation of the CEO and Chair roles because they claim that separation promotes board independence and accountability (Del Guercio and Hawkins, 1999; USA Today, 2004; The Economist, 2002; Wall Street Journal, 2010). Similarly, the Cadbury Report (Cadbury, 1992), a governance reform document written in the United Kingdom favors separating the two top positions. In contrast, practitioners have long maintained some firms benefit from combining the two roles. While not mandating the separation of the two positions in publicly held corporations, the Securities and Exchange Commission (SEC) proposed new disclosure requirements mandating companies to disclose more information about their decision to combine or separate the two positions. 1 Some academic studies posit that it is beneficial for firms to separate the two roles (e.g., Fama and Jensen, 1983; Rechner and Dalton, 1991), while other studies find that combining the roles does not lead to lower firm performance (e.g., Daily and Dalton, 1992; Brickley et al., 1997). Collectively these studies suggest that there may be both costs and benefits to the CEO also being the Chair. While some settings favor a separation of the two roles, other settings favor a combination. 2 Consequently, the first part of our study examines whether and how a firm s operating environment plays a critical role in affecting the benefits and costs of the CEO also serving as the Chair. 1 According to a proposal of a new disclosure requirement to Item 407 of Regulation S-K, and as a corresponding amendment to Item 7 of Schedule 14A, the SEC (2009) requires public companies to: (1) disclose whether and why they have chosen to separate or combine the CEO and Chair position, (2) disclose whether and why firms have a lead independent director when the two roles are combined, and (3) specify the role the lead independent director plays in the leadership of the company. 2 This notion is also stated in the SEC s proxy disclosure proposals pertaining to this issue, where the SEC clearly states that it does not advocate either separation or combination of these roles but leaves it to the companies to decide which is suited for their needs ( 2

3 Much of the governance research focuses on the monitoring role of the board but not on its advisory role. 3 In this study, we hypothesize that firms with greater advisory needs or those that require more managerial initiatives for innovation are more likely to combine the CEO and Chair positions. We contend that when management is more reliant on board members for their advice on various corporate matters, it is beneficial for firms to have the CEO also serve as the Chair. Management may need to rely on the expertise of board members when facing complex corporate decisions or when the firm faces adverse operational or financial situations (Coles, Daniel, and Naveen, 2008). Compared to a CEO who is not a Chair, a CEO-Chair has considerably more discretion in board matters (Adams, Almeida, and Ferreira, 2005) including setting the agenda for board meetings, hiring board members with necessary expertise, and nominating various sub-committees where the CEO needs the best advice. Since the CEO has intimate and specialized knowledge of the firm s operation and financial situation, allowing the CEO to be the Chair facilitates better information flows between management and the board. Having separate CEO and Chair roles for a firm with greater advisory needs may prove to be especially costly if conflicts and rivalry between the CEO and Chair prevent such information flows in a timely manner (Donaldson and Davis, 1991; Finkelstein and D Aveni, 1994; Brickley et al., 1997). Similarly, firms relying on more managerial initiatives for innovation (e.g., growth firms) may need to grant CEOs greater latitude in pursuing activities they deem fit without imposing added restrictions from board members (Burkart et al., 1997; Boot et al. 2006). CEOs in such firms need greater discretion as they search for profitable new investment projects (Burkart et al., 1997), or as they invest in R&D to remain competitive than CEOs in firms where fewer such initiatives for innovation are necessary (Prendergast, 2002; Christie et al., 2003). A manager is less likely to show such initiatives when there are added restrictions imposed on his/her discretion. Because CEOs are less restricted in their activities when they also hold dual positions, they have greater latitude in making timely decisions (Adams et al., 2005). 3 Only a handful of studies examine the advisory role of boards. For example, Helland and Sykuta (2004) provide evidence that a more regulated external environment surrounding the firm increases the proportion of directors who are affiliated with political regulatory environment. Similarly, Agrawal and Knoeber (2001) find that politically experienced outside directors are more prevalent on boards where politics is an important determinant of firm profitability. 3

4 Further, when a firm has a separate Chair, necessary information transfer may be delayed between the CEO and the Chair due to rivalry between these two individuals, busy schedules, or other reasons (Brickley et al., 1997). These costs could be especially large when firms depend on their CEOs to take timely innovative actions. Thus, firms relying on more managerial initiatives for innovation are more likely to appoint CEOs as Chairs. Like Coles et al. (2008), we use firm size, diversification, and leverage to proxy for firms advisory needs. We use market-to-book ratio, research and development intensity, advertising intensity, and intangible intensity as constructs for managerial initiatives for innovation (Lev, 1983; Smith and Watts, 1992; Gaver and Gaver, 1993). Advisory needs and managerial initiatives for innovation are unobservable and any single proxy is likely to measure the two constructs with error. Therefore, as in prior studies (e.g., Gaver and Gaver, 1993; Coles et al., 2008; Garfinkel 2009), we combine the information from the multiple proxies using factor analyses. Factor analysis enables us to reduce the number of variables into two distinct factors that capture advisory needs and managerial initiatives for innovation. Consistent with our prediction, we find that the likelihood a CEO also serves as the Chair is higher when a firm has greater advisory needs. Our findings are robust to including several CEO- and firm-level variables and industry fixed effects. However, we only find weak evidence supporting our prediction of a positive relation between a firm s need for managerial initiatives for innovation and the likelihood that the CEO is also Chair. More important, in the second part of our study, we also examine whether CEOs who are Chairs extract rents for themselves at the cost of shareholders. One concern is that, while the decision to appoint a CEO may be optimal, CEOs may subsequently decide to use the added power associated with the dual position to extract rents for themselves. Because CEO-Chairs have greater discretion in decision-making, in selecting and firing board members, and in setting the agenda for board meetings compared to CEOs who are not Chairs, they are expected to yield greater power. According to the managerial power hypothesis, CEOs are more likely to use their added power from holding dual positions to extract rents (Bebchuk, Fried, and Walker, 2002). In contrast, optimal contracting suggests that CEO-Chairs are unable 4

5 to extract rents. While CEOs have significantly more power under dual leadership structures, a stronger governance mechanism in such firms prevents CEOs from misusing their power to enhance personal welfare. Thus, stronger governance is optimal for firms with dual leadership structure because it exploits the potential benefits of having CEO-Chairs while minimizing related costs. By examining the relation between a CEO being the Chair and the likelihood of rent extraction, we add to the debate on whether rent extraction or optimal contracting is more pervasive. CEO-Chairs could extract rents by managing earnings to receive greater compensation (Healy, 1985), to inflate stock price of the firm (Barth et al., 1999; Ghosh et al., 2005), or to affect their career prospects by influencing capital market perceptions (Bartov et al., 2002). They could also use their influence over board members to negotiate excessively high compensation or to place greater reliance on earnings- relative to stock-based performance measures in determining their compensation. 4 Finally, lower CEO turnover when CEOs are also Chairs could signal CEO entrenchment because they are able to stay longer than other CEOs by using their influence over boards. Consistent with prior research, we use the following variables to examine whether CEOs with dual positions extract rents by: (1) using accounting accruals, and real activities to manage earnings, which might eventually lead to a financial restatement (Dechow et al., 1996; Dechow and Dichev, 2002; Roychowdhury, 2006; Hennes et al., 2008), (2) paying themselves excessive compensation (Core et al., 2008), (3) ensuring that their pay is more sensitive to earnings than to stock returns (Davila and Penalva, 2006; Cheng and Indjejikian, 2009), and (4) ensuring longer tenure through less frequent turnover (Weisbach, 1988). Since the decision to appoint the CEO also as a Chair depends on the firm s operating environment, and is therefore not exogenous, we use a two-stage least squares regression model to examine whether CEO-Chairs extract rents. In the first stage, we regress an indicator variable for CEO- Chair on advisory and managerial initiatives for innovation factors while including other control 4 Although prior studies (e.g., Lambert and Larcker, 1987; Sloan, 1993) show that accounting earnings and stock returns are both used to determine CEO pay, stock returns are more directly associated with maximizing shareholder welfare than accounting earnings. Also, prior studies suggest that managers might prefer accounting-based performance evaluation measures (like earnings) over market-based measures which are harder to influence (e.g., Ittner, Larcker, and Rajan, 1997; Davila and Penalva, 2006; Cheng and Indjejikian, 2009). 5

6 variables. In the second stage, we regress our constructs for rent extraction on the predicted values of CEO-Chair from the first stage regression. In general, we find no evidence of CEO-Chairs manipulating earnings using accruals quality, discretionary accruals, real activities, or that they are more likely to violate GAAP using financial restatements as a measure of GAAP violation. 5 Further, we find no evidence that pay-earnings sensitivity is greater than pay-stock returns sensitivity for CEO-Chairs, indicating that CEO-Chairs do not rely on performance measures over which they have more control to influence their compensation indirectly. In contrast, we do find that CEO-Chairs receive excessive compensation and that turnover is lower when they hold dual positions, which suggests some evidence of rent extraction. Our findings suggest that there are different dimensions to rent extraction; CEO-Chairs do not extract indirect rents through earnings management or pay-performance sensitivity, rather, they extract direct rents in the form of higher pay and greater job security. One implication of this finding is that CEO-Chairs act strategically in their rent extraction. While earnings management might lead to greater pay for CEOs if a larger portion of their compensation is tied to bonuses, or it might increase their stock or option values if stock prices respond more favorably to reported earnings, the personal rents to CEOs from earnings management are uncertain. Instead, CEOs appear to use their added power to negotiate for more direct personal benefits including more favorable compensation contracts and protection from job losses which in turn yield greater future compensation. One could argue that higher pay and greater job security are optimal because CEO-Chairs have special managerial talent, which merits higher pay and job security. While our regression analyses control for CEO- and firm-specific factors that are expected to capture managerial ability, it is possible that we fail to adequately proxy for CEO s managerial ability in our regressions. Therefore, we estimate our regressions after including CEO fixed effects, which are constant over time but vary across firms. As in 5 We examine different possibilities for earnings/real activities management that have been documented in the literature (e.g., Dechow and Dichev, 2002; Roychowdhury, 2006) as managers could make tradeoffs between these types of opportunistic actions to extract rents. For example, some managers could place more emphasis on real activities management as this is less detectable than accruals management. 6

7 Bamber et al. (2010), if CEO-Chairs have special managerial talents, manager specific fixed effects are expected to capture CEO specific attributes and abilities. The results using CEO-specific fixed effects are again consistent with the rent extraction perspective and inconsistent with the notion that the added pay and job security reflect some unobservable managerial ability. In summary, our results suggest firms and CEOs both benefit when CEOs are appointed as the Chair. Firms that are more reliant on their boards for expert advice benefit from having CEOs also serve as the Chair. Similarly, firms relying on more managerial initiatives for innovation benefit from giving the CEO more power and latitude to make decisions without hindrance from the Chair of the Board. However, it also appears that CEOs personal benefits are also substantial when they serve in their dual capacity due to receiving excess compensation and greater job security. We contribute to the literature in at least three ways. First, by showing that greater advisory and managerial initiatives for innovation needs generally lead to a higher likelihood of the CEO being Chair, we provide an explanation for the observed heterogeneity in the CEO-Chair positions. Our results suggest that a one-size-fits-all approach to corporate governance as proposed by some studies may not be efficient. Second, we contribute to the debate on optimal contracting versus managerial power with respect to CEO-Chairs. Our results suggest that researchers should consider an array of rent extraction variables rather than focus on one or two variables as in some prior studies (e.g., Core et al., 1999; Bebchuk et al., 2002). Because our results are consistent with both optimal contracting and rent extraction, it is understandable why the topic of dual leadership structure generates so much controversy. Finally, our study contributes to the accounting literature that links governance with earnings/real activities quality variables by adding to the existing literature that governance is not related to earnings management. The next section contains the theoretical background for this study leading to the hypotheses. Section 3 outlines the research design, and section 4 reports the sample description. Section 5 discusses the empirical results, and section 6 concludes our study. 7

8 2. Theoretical Background In this study, we hypothesize that the decision to separate or combine the CEO and Chair roles is determined after netting the costs and benefits of the two decisions. Over the past 20 years, numerous studies examine whether separating or combining the CEO-Chair positions impacts firm performance. Some studies contend that the two roles should be separated because independent CEO oversight fosters managerial accountability, which reduces agency conflicts between the CEO and shareholders and therefore improves firm performance (e.g., Fama and Jensen, 1983; Rechner and Dalton, 1991; Jensen, 1993). In contrast, other studies find no evidence indicating that combining these roles is associated with inferior firm performance (e.g., Daily and Dalton, 1992; Brickley et al., 1997). These studies generally contend that information sharing between the CEO and a non-executive chairman is difficult when the CEO has proprietary information due to rivalry between the two individuals. Furthermore, it could be costly for the CEO to communicate timely information to a non-executive chairman because of conflicts between the two individuals. The CEO may also not have the necessary flexibility or authority in some key areas when there is a separate Chair. For example, a CEO might form sub-committees to advise him/her about certain profitable but risky investment projects which may require further approvals from a non-executive Chair before being implemented. To summarize, benefits of separating the two roles include a more independent board that is not beholden to the CEO, and a more accountable CEO. Some benefits of combining the two roles include fewer board room deliberations, CEO-Chair being able to take timely actions, and more sharing of proprietary information with board members. Our central hypothesis is that the decision to combine the CEO and Chair roles depends on whether a firm has greater advisory needs and whether more managerial initiatives for innovation are crucial for the firm, which ultimately determines whether it is optimal to have the CEO also serve as the Chair. In the following subsections, we discuss the reasoning behind our related hypotheses. 8

9 Firms with Advisory Needs Some firms have a greater need for advice from board members than other firms. For example, larger firms, more diversified firms, or firm with more complex operations benefit from seeking the advice of board members on complex matters (Klein, 1998). Because the quality of advice depends on the CEO s willingness to share proprietary information about the firm s operations and financial position, a CEO seeking advice from its board is likely to be more forthcoming and to share more information with the board. We posit that whether the CEO is Chair or not affects the flow of material information between the management and board. As the top executive of the firm, CEOs have unparalleled specialized knowledge about the operations, strategic opportunities and challenges facing the firm. CEOs are more likely to share such information when they also serve as Chairs for several reasons. First, CEO-Chairs have significant discretion in hiring board members (Adams et al., 2005), especially those they feel are ideally suited to offer expert advice. A CEO is more likely to share information with those he/she helped recruit personally. Second, because a CEO-chair also has the discretion in calling a meeting and in setting the agenda (Florou, 2005), a CEO-Chair is ideally positioned to expedite the flow of information and then seek the necessary feedback from the board. In contrast, when the CEO is not the Chair, due to rivalry and conflicts between the CEO and non-executive Chair, the CEO may not be as willing or as able to share proprietary information with the board. Additionally, any confusion about the power structure may impede the flow of information from other board members to the CEO (Lipton and Lorsch, 1992; Finkelstein and D Aveni, 1994). These types of frictions between the CEO and non-executive Chair are likely to increase agency costs, which are costly for a firm with greater advisory needs. The preceding discussion leads to this study s first hypothesis (stated in alternate form): H1: Firms with greater advisory needs are more likely to have a CEO also serve as the Chair of the board of directors, ceteris paribus. 9

10 Firms Relying on Managerial Initiatives for Innovation When firms rely on more managerial initiatives for innovation, CEOs need to be given greater discretion so that they can make decisions without restrictions (Boot, Gopalan, and Thakor, 2006; Burkart, Gromb, and Panunzi, 1997). 6 With greater managerial discretion, CEOs are free to pursue new investment projects, invest in R&D, undertake necessary acquisitions, engage in differentiation activities, and use their specialized knowledge without having to worry about the board s intervention on major investment projects (Prendergast, 2002; Christie, Joye, and Watts, 2003). CEOs are unlikely or unable to show such initiatives when there are restrictions imposed on their activities. CEOs who are also Chairs have greater discretion in decision-making compared to CEOs who are not Chairs (Adams et al., 2005). When firms rely on more managerial innovations, they are more likely to appoint the CEOs as Chair of the board of directors to encourage them to use their discretion to make investment decisions while imposing fewer restrictions. Because managers have greater discretion in decision-making when they serve dual positions, CEO-Chairs have the flexibility and authority to engage in innovative pursuits as they deem necessary to remain competitive. Thus, according to our second hypothesis (stated in alternate form): H2: Firms that require more managerial initiatives for innovation are more likely to have a CEO also serve as a Chair of the board of directors, ceteris paribus. CEO-Chairs and Rent Extraction Since CEO-Chairs have greater discretion in decision-making, selecting and firing board members, and setting the agenda for board meetings compared to CEOs who are not Chairs, they are expected to yield greater power (Brickley et al., 1997). Also, when the CEO is Chair, the CEO is generally less accountable to the board. Furthermore, board members are less likely to challenge the CEO s decisions as many of them, if not all, are likely to have been selected by the CEO. Thus, a typical CEO-Chair is expected to have significant influence over the board than a CEO who is not a Chair. 6 Innovation is defined here as the successful implementation of creative ideas to compete and differentiate in the marketplace (Amabile et al., 1996; Baregheh et al., 2009). 10

11 According to the optimal contracting perspective, contractual arrangements and other mechanisms in firms are optimally designed to maximize shareholder welfare. Under this perspective, even when managers have significant power, as in the case of CEO-Chair, they have limited scope or incentives to extract rents for themselves while eroding shareholder value because other contracting or monitoring mechanisms exist to prevent such rent extraction at the expense of shareholders. For example, boards might be structured to have higher levels of independence, effectiveness and diligence when appointing the CEO also as the Chair. CEOs appropriating rents could be reprimanded, and in extreme cases, even be fired by a performing board. Similarly, CEO-Chairs might be associated with firms when there are activist shareholders or other blockholders who would ensure through vote no campaigns and proxy fights that ineffective CEO-Chairs are removed from the office. Also, one prerequisite for being a CEO-Chair might be to require a CEO-Chair to have significant ownership in the firm. When CEO-Chairs use their power to benefit personally while eroding shareholder value, their actions impose a large cost on them because of their ownership in the firm. Thus, with significant ownership, the cost of rent extraction becomes larger than the benefit and, consequently, CEO-Chairs are less likely to use the power to extract personal rents. In contrast, according to the managerial power/rent extraction perspective, managers are able to exercise their power to extract rents for themselves while harming shareholders (Bebchuk, Fried, and Walker, 2002). This view departs from optimal contracting because a corporate board is more beholden to CEO-Chairs and therefore unable to check their power and any rent extraction. 7 Because the two hypotheses predict conflicting outcomes, the influence of CEO-Chairs on rent extraction is ambiguous. Therefore, according to the managerial power (optimal contracting) hypothesis: H3: CEOs who are also Chairs of their boards (do not) use their added power from holding dual positions to extract rents from their firms. 7 The evidence is mixed on whether the optimal contracting or managerial power/rent extraction perspective prevails empirically. Some studies find evidence in support of the optimal contracting but not the managerial power/rent extraction perspective (e.g., Bowen et al., 2008) whereas other studies find evidence that managers use their power to extract rents from their firms (e.g., Bebchuk et al., 2002). We attempt to shed some light on this debate by examining whether CEOs who are also Chairs use their added power to extract rents. 11

12 In this section, we discuss how the likelihood of the CEO being Chair of the board of directors is determined by a need for a greater advisory environment in firms and also an environment where more managerial initiatives for innovation are necessary. We also discuss the implications of the CEO being Chair on the likelihood of rent extraction in firms. In the next section, we discuss our research methodology. 3. Research Methodology Our analysis comprises of two parts. To test our first two hypotheses, where we predict that firms with greater advisory and managerial initiatives for innovation needs are more likely to appoint their CEOs as Chairs, we regress the likelihood that the CEO is Chair on variables that proxy for advisory needs and managerial initiatives for innovation needs and a range of control variables. Because the decision to appoint a CEO as Chair is endogenously determined, we use the two-stage least squares regression methodology to test our third hypothesis which, in its null form, predicts that CEO-Chairs do not use their power to extract rents. We use the predicted values for the CEO-Chair variable obtained from the first stage in the second stage regression where we regress a range of variables that capture rent extraction on the predicted CEO-Chair and other control variables Firms advisory and managerial initiatives for innovation needs and the likelihood of the CEO being Chair To test our first two hypotheses, we run the following logistic regression model: Prob(CEO-Chair = 1) = Advice + 2 Initiative + 3 Governance + 4 ROA + 5 StockReturn + 6 Incentive-Compensation + 7 Firm-Age + 8 CEO-Ownership + 9 CEO-Tenure + 10 CEO-Age + 11 Industry dummy + 12 Year dummy + (1) The regression model in Equation (1) analyzes whether the firms advisory needs and managerial initiatives for innovation needs affect the likelihood of the CEO being the Chair. The variables used in the 12

13 regression are defined in the appendix. CEO-Chair, the dependent variable, is an indicator variable that equals one if the CEO is Chair and zero otherwise. Similar to Coles et al. (2008), to construct Advice, we first conduct a factor analysis of several variables that proxy for a firm s advisory needs, including firm size, industrial diversification, and financial leverage, which capture complexities within a firm and therefore require greater advisory needs. Using a minimum Eigenvalue of one as the selection criterion for eliciting a factor, we extract one factor that represents firm complexity and compute a factor score based on a linear combination of the three variables. Firm size (Sales) is measured by the log of firm sales; industrial diversification (Segments) is measured by the number of business segments in which the firm operates; and financial leverage (Leverage) is measured by the book value of total debt divided by the book value of total assets. As in Coles et al. (2008), we then construct an indicator variable Advice that equals one when the factor score obtained by conducting the factor analysis is greater than its median value and zero otherwise. Initiative is the variable we use to measure firms managerial initiatives for innovation needs. To construct Initiative, we first conduct a factor analysis of several variables that proxy for a firm s managerial initiatives for innovation including market-to-book ratio, research and development (R&D) intensity, advertising intensity, and intangible intensity (Lev, 1983; Smith and Watts, 1992; Gaver and Gaver, 1993). Similar to computing the factor score for advisory needs, we obtain one factor with inputs being the four proxies for managerial initiatives for innovation needs. Again, only one factor is significant and we compute a factor score based on these four variables. The market-to-book ratio (MTB_Equity) is measured by dividing the market value of equity by the book value of equity; R&D intensity (RD-Exp) and advertising intensity (AD-Exp) are measured by dividing research and development and advertising expenditure by sales, respectively; and intangible intensity (Intangible) is measured as follows: 1 minus ratio of net property, plant, and equipment to book value of total assets. We then construct Initiative as an indicator variable that equals one when the factor score obtained by conducting the factor analysis is greater than its median value and zero otherwise. 13

14 The rest of the independent variables in Equation (1) are control variables. Governance is a variable that controls for overall governance in a firm. Thus, it controls for monitoring explanations for the likelihood that the CEO is Chair, which has dominated governance research to date (e.g., Brickley et al., 1997). Similar to Gillan et al. (2003), we construct Governance by computing an average percentile of several variables (ranked in order of greater monitoring) that are commonly used in the literature to denote governance/monitoring intensity (e.g., Yermack, 1996; Gompers et al., 2003). These variables are board size, board independence, institutional ownership, and the extent of shareholder rights. Board size (Board-Size) is the number of directors on the board and board independence (Board-Independence) is the percentage of outside (non-affiliated) directors on the board; institutional ownership (INST-Ownership) is measured by the percentage of outstanding common stock held by financial institutions at the fiscal yearend; and the extent of shareholder rights (G-Index) is measured by the G-index (Gompers et al., 2003), which measures the strength of shareholder rights based on 24 charter provisions and by-laws. Consistent with much prior research, we assume that smaller boards, more outsiders on the board, greater institutional ownership, and greater shareholder rights are associated with more intense monitoring (e.g., Weisbach, 1988; Yermack, 1996; Gompers et al., 2003; Hartzell and Starks, 2003). Larger values of Governance are associated with stronger monitoring. ROA is the ratio of earnings before interest, taxes, depreciation, and amortization to average book value of assets. StockReturn is the buy-and-hold annual stock returns. Incentive-Compensation is the sum of the CEO s bonus, the Black-Scholes value of stock options granted, the value of restricted stock grants, and long-term incentive payout scaled by total compensation during the year. Firm-Age is the number of years that the firm is publicly traded as of the fiscal year-end. CEO-Ownership is measured by the percentage of outstanding common stock held by the CEO at the fiscal year-end. CEO-Tenure is the number of years that the CEO has been in office as of the firm s fiscal year-end. Finally, CEO-Age is the CEO s age at the end of fiscal year. 14

15 Our first hypothesis predicts that when firms have greater advisory needs, they are more likely to appoint their CEOs as Chairs. To test this, we examine the coefficient on Advice, 1, in Equation (1). Our first hypothesis is supported if 1 > 0. Our second hypothesis predicts that when firms have more managerial initiatives for innovation needs, they are more likely to appoint their CEOs as Chairs. To test this, we examine the coefficient on Initiative, 2, in Equation (1). Our second hypothesis is supported if 2 > CEO-Chair and the likelihood of rent extraction To determine whether CEO-Chairs use their added power from holding dual positions to extract rents, we use a range of variables that have been used in prior research to measure rent extraction. Specifically, we examine whether: (a) CEO-Chairs are more likely to engage in earnings management than other CEOs (Jones, 1991; Dechow and Dichev, 2002; Roychowdhury, 2006; Hennes et al., 2008), (b) CEO-Chairs earn excessive compensation (Core et al., 2008), (c) CEO-Chairs have greater payearnings sensitivity compared to pay-stock returns sensitivity than other CEOs (Davila and Penalva, 2006; Cheng and Indjejikian, 2009), and (d) CEO-Chairs have greater job retention than other CEOs (Weisbach, 1988). In our tests, we regress our rent extraction variables on the predicted values of the CEO-Chair variable from Equation (1) to test for specific instances of rent extraction. (a) CEO-Chair and earnings management Prior research has shown that managers manipulate earnings to maximize their compensation (Healy, 1985), to influence the stock price (Barth et al., 1999; Ghosh et al., 2005), or to affect their career prospects by influencing capital market perceptions (Bartov et al., 2002). To examine whether CEO- Chairs manage earnings, we regress our earnings management variable on predicted CEO-Chair (CEO- Chair Pred ) and a range of control variables as follows: Earnings management = CEO-Chair Pred + 2 Governance + 3 ln(mtb_equity) + 4 Operating-Cycle + 5 Volatility CFO + 6 Volatility Sales + 7 Foreign + 8 ln(segments) + 15

16 9 Incentive-Compensation + 10 Firm-Age + 11 Large-Auditor + 12 ln(mve) + 13 Leverage + 14 Litigation + 15 Loss + (2) We use several proxies for earnings management which lead to different sets of regressions. Our first proxy for earnings management is accruals quality (Dechow and Dichev, 2002). Accruals quality is estimated based on the Dechow-Dichev (2002) model. We employ two alternative measures of accruals quality measured over the sample period. Accruals-Variance1 is the standard deviation of the firm i s residuals from annual firm-level estimations of WC t = β 0 + β 1 CFO t-1 + β 2 CFO t + β 3 CFO t+1 + ε t, where WC is the change in working capital, CFO is operating cash flows, and all the variables are scaled by average total assets. Accruals-Variance1 is computed based on the assumption in Dechow and Dichev (2002) that the accounting cycle from payments to cash receipts occurs within a year. To relax this assumption, we construct another measure of accruals quality, Accruals-Variance2, taking into consideration that accruals need more than one year to be realized. Accruals-Variance2 is the standard deviation of firm i s residuals from annual firm-level estimations of WC t = β 0 + β 1 CFO t-2 + β 2 CFO t-1 + β 3 CFO t + β 4 CFO t+1 + β 5 CFO t+2 + ε t. Our second set of earnings management variables measures discretionary accruals based on the cross-sectional modified Jones model (Jones, 1991). We employ three alternative measures of discretionary accruals: (1) Discretionary-Accruals1 is the absolute discretionary accruals estimated from the cross-sectional modified Jones model, (2) Discretionary-Accruals2 is the absolute discretionary accruals estimated from the cross-sectional modified Jones model with ROA in the accruals regression to control for firm performance, and (3) Discretionary-Accruals3 is the absolute portfolio performance adjusted discretionary accruals, where portfolio performance adjusted discretionary accruals is the difference between a firm s discretionary accruals and the median discretionary accruals for the matching ROA portfolio. 8 Our third set of earnings management variables measures real activities management (Roychowdhury, 2006). Real activities management is estimated from the cross-sectional Roychowdhury 8 Like Kothari et al. (2005), we adjust for firm performance to control for the effect of performance on accruals. 16

17 model. Again, we employ three alternative measures of real activities management: (1) Activity- Management1 is the absolute abnormal cash flow estimated from the cross-sectional Roychowdhury model, where abnormal cash flow from operations is the actual cash flow minus the normal cash flow calculated using estimated coefficients from the corresponding industry-year model and the firm year s sales and lagged assets, (2) Activity-Management2 is the absolute abnormal discretionary expenses estimated from the cross-sectional Roychowdhury model where discretionary expenses are defined as the sum of R&D, advertising, and selling, general and administrative expenses, and (3) Activity- Management3 is the absolute abnormal production costs estimated from the cross-sectional Roychowdhury model where production costs are defined as the sum of cost of goods sold and change in inventory. Our final set of earnings management variables includes financial restatements because of accounting irregularities (Hennes et al., 2008). Although firms restate earnings due to reasons like changes in accounting policies, restatements that are irregularities constitute clear and material violation of appropriate accounting practices as they represent a direct admission of misrepresentation (Agrawal and Chadha, 2005). We employ an OLS regression model when we use our first three sets of earnings management variables. However, when we use restatements, we employ a logistic regression model to examine the likelihood of firms restating earnings when the CEO is Chair. In our regression, Restatement equals one when a firm restates its financial statements due to accounting irregularities during the sample period and zero otherwise. We include several control variables in all our regressions. Operating-Cycle is the sum of the number of days of accounts receivable and inventory outstanding; Volatility CFO and Volatility Sales are the standard deviation of cash flows (divided by lagged total assets) and cash-based revenues (divided by lagged total assets), respectively; Foreign is an indicator variable that equals one if foreign currency adjustment is non-zero and zero otherwise; ln(segments) is the natural logarithm of one plus the number of operational and business segments; Large-Auditor is an indicator variable that equals one when the client s auditor is a large accounting firm and zero otherwise; Litigation is an indicator variable that equals one if the firm operates in a high-litigation industry and zero otherwise (high-litigation industries 17

18 are industries with four-digit SIC codes of , , , , and ); and Loss is an indicator variable that equals one if the firm reports a net loss and zero otherwise. For the regression where Restatement is the dependent variable, we include an additional control variable, Cash-Flow, which measures cash flow from operations scaled by beginning of the year total assets. All other variables are as previously defined. All explanatory variables are measured as the average of the annual values over the sample period. In all our regressions, to examine whether CEO-chairs are more likely to manage earnings, we examine the coefficient on CEO-Chair Pred, 1. If this coefficient is positive, then this is evidence that CEO-Chairs extract rents by managing earnings. (b) CEO-Chair and excess compensation If managers use their power to influence their compensation, they are likely to receive compensation in excess of what is optimal as determined by a range of economic attributes of the CEO, firm, and industry. To examine whether CEO-Chairs receive excessive compensation, we first estimate the following OLS regression: Compensation = ROA+ 2 StockReturn + 3 CEO-Ownership + 4 CEO-Tenure + 5 CEO- Age + 6 Market-Return + 7 Volatility Return + 8 ln(mve) + 9 ln(mtb_equity)+ 10 ln(ceo-option-hold) + 11 ln(ceo-stock-hold) + 12 Industry dummy + 13 Year dummy + (3) As in Core et al. (2008), we measure Excess-Compensation by subtracting expected compensation from the logarithm of compensation where expected compensation is estimated as the predicted value of Compensation from Equation (3). In Equation (3), we use three different flow compensation variables: (1) Salary is the logarithmic transformation of the amount of salary received for the fiscal year, (2) Bonus is the logarithmic transformation of the amount of bonus received for the fiscal year, and (3) Stock is the logarithmic transformation of the sum of the Black-Scholes value of stock options granted and the value of restricted stock grants during the year. 18

19 The control variables included in the compensation regression are defined as follows: Market- Return is the return on the S&P 500 index; Volatility Return is the standard deviation of StockReturn over the previous sixty months and is a proxy for firm risk; ln(mve) is the logarithm of the market value of equity and is a proxy for firm size; ln(ceo-option-hold) is the logarithm of the dollar value of CEO option holdings at the end of the prior year; and ln(ceo-stock-hold) is the logarithm of the dollar value of CEO stock holdings (including restricted stock) at the end of the prior year. All the other variables are as previously defined. We measure the relation between Excess-Compensation and CEO-Chair using the following regression. Excess-Compensation = CEO-Chair Pred + 2 Governance + υ (4) If CEO-Chairs extract rents via excessive compensation, then the coefficient on CEO-Chair Pred, 1, in Equation (4) is positive because it indicates that CEOs use their power to increase their compensation beyond optimal levels that are based on the firm s operating and information environment, and CEO and industry factors. (c) CEO-Chair and pay-earnings sensitivity relative to pay-stock returns sensitivity Prior research has shown that firms rely on both earnings- and stock-based performance measures to determine managers incentive compensation (Lambert and Larcker, 1987; Sloan, 1993). Managers better understand how their actions map into earnings than the stock price (Gjesdal, 1981; Murphy, 1999). While stock returns measure shareholder economic value, they are a noisier measure of managerial actions as they reflect the market s expectations of future cash flows. Davila and Penalva (2006) and Cheng and Indjejikian (2009) show that when managers have power, they tend to extract rents by negotiating that a greater proportion of their incentive compensation is based on earnings-based measures than on stock-based measures. Hence, CEO-Chairs with considerable power and intending to extract rents 19

20 would negotiate for a heavier weight on earnings relative to stock returns in determining their compensation. To determine this, we estimate the following regression: Compensation = CEO-Chair Pred + 2 Governance + 3 CEO-Chair Pred ROA + 4 CEO- Chair Pred StockReturn + 5 ROA + 6 StockReturn + 7 CEO-Ownership + 8 CEO- Tenure + 9 CEO-Age + 10 Market-Return + 11 Volatility ROA + 12 Volatility Return + 13 ln(mve) + 14 ln(mtb_equity)+ 15 ln(ceo-option-hold) + 16 ln(ceo-stock-hold) + 17 Industry dummy + 18 Year dummy + (5) where Compensation is the first difference in the logarithm of compensation between the current year and the prior year. We use two different specifications for our compensation variable. The first is cash compensation, which is the sum of salary and bonus. The second is total compensation, which is the sum of salary, bonus, long-term incentive plan payouts, the value of restricted stock grants, and the Black- Scholes value of stock options granted during the year. Changes in CEO compensation are computed when the current CEO was also the CEO in the previous year. ROA is the first difference in ROA. Volatility ROA is the standard deviation of ROA over the previous five years. All other variables are as defined above. To determine whether CEO-Chairs extract rents from their firms by negotiating for greater payearnings sensitivity compared to pay-stock returns sensitivity, we compare the coefficient, 3, which measures the pay-earnings sensitivity, to 4, which measures pay-stock returns sensitivity, when the CEO is Chair. 9 (d) CEO-Chair and job retention Our final test of whether CEO-Chairs extract rents comes in the form of job retention. It is likely that when managers have added power, as in the case of CEOs who are also Chairs, they are able to 9 We first determine whether these coefficients are significant. If they are, then additional modifications and tests are required to more precisely examine the nature of the pay-performance relations. For example, the performance measures may need to be standardized to account for scaling differences between the two variables. 20

21 bargain for greater job security. Hence, if CEO-Chairs use their added power to be entrenched in their firms, we would expect CEO-chairs to have lower turnover than other CEOs. 10 To test this, we run the following logistic regression: CEO-Turnover = CEO-Chair Pred + 2 Governance + 3 CEO-Chair Pred ROA + 4 CEO- Chair Pred StockReturn + 5 ROA + 6 StockReturn + 7 CEO-Ownership + 8 CEO- Tenure + 9 CEO-Age + 10 Market-Return + 11 Volatility ROA + 12 Volatility Return + 13 ln(mve) + 14 ln(mtb_equity)+ 15 ln(ceo-option-hold)+ 16 ln(ceo-stock-hold) + 17 Industry dummy + 18 Year dummy + (6) where CEO-Turnover equals one if a CEO change occurs during the fiscal year and zero otherwise. All other variables are as defined above. If CEO-Chairs extract rents by bargaining for greater job security, then we expect 1 to be negative. In a recent study, Gow et al. (2010) compare different methods that address the issue of crosssectional and time-series dependence of the residuals for panel data sets. They find that clustered standard errors from pooled regressions provide unbiased estimates of the true standard errors, whereas Newey- West and Fama-MacBeth standard errors are biased. Specifically, Gow et al. (2010) recommend a twolayer clustering (a) by time to allow for cross-sectional dependence of the residuals and (b) by firm to allow for time-series dependence of the residuals. Where possible, we follow this procedure and report the statistical significance of the estimated coefficients based on clustered standard errors throughout our analyses. In summary, we first test whether firms with greater advisory and managerial initiatives for innovation needs are more likely to appoint their CEOs as Chairs. We then include several tests to examine whether CEO-chairs extract rents from their firms by using their added power using a range of rent extraction methods. 10 It is arguable that higher CEO turnover could be due to the CEO being wooed away by rival firms. However, this would only add noise to our analysis and bias us against finding a result. 21

22 4. Data and Descriptive Statistics 4.1. Sample selection Our initial sample consists of all firms with information on corporate governance from the RiskMetrics database (previously known as IRRC) with annual shareholder meetings held between 1998 and RiskMetrics examines the proxy statements for firms listed in the Standard and Poor s (S&P) 500 index, the S&P MidCap 400 index, and the S&P SmallCap 600 index. We obtain data on board characteristics from RiskMetrics. Since the RiskMetrics data are based on the proxy meeting year, we obtain the fiscal year corresponding to the proxy meeting year by comparing the fiscal year-end to the meeting date. Further, we obtain institutional stock holdings data from CDA/Spectrum gathered from 13F forms filed with the SEC. We also collect the G-index made available through the website of Andrew Metrick. 11 We obtain CEO compensation, stock ownership, tenure, and age data from Compustat s ExecuComp database. The data on firm characteristics are collected from the Compustat annual files. Stock return and firm age data are obtained from CRSP. We then merge the governance data with the CEO, accounting, and stock return data. We eliminate 69 firm-year observations with negative values of MTB_Equity, as in Barth et al. (1999), to mitigate effects of troubled firms with negative net worth. The raw distributions of several variables indicate severe outliers and, thus, we winsorize the top or bottom 1 percent of the observations for Leverage, MTB_Equity, RD-Exp, AD-Exp, ROA, and StockReturn to reduce the impact of influential observations. This sample selection procedure results in 5,659 firm-year observations over fiscal years 1998 through Descriptive statistics Table 1 provides the descriptive statistics for the variables used in Equation (1). The majority of the boards have the same CEO and Chair (CEO-Chair); the CEO is also the board Chair in about 67% of 11 Andrew Metrick, Governance Index Data ( Although the G- index is available only for 1998, 2000, 2002, 2004, and 2006, it is rather stable over time. We follow the standard practice of using the closest prior measure (lagging 1 year) if the G-index is not available for a firm year (e.g., Gompers et al., 2003; Core et al., 2006). 22