CEO Compensation and Board Oversight

CEO Compensation and Board Oversight Vidhi Chhaochharia Yaniv Grinstein ** Preliminary and incomplete Comments welcome Please do not quote without permission In response to the corporate scandals in 2001-2002, the major U.S. exchanges came up with new requirements from corporate boards which are intended to increase board oversight. We use this regulation event to shed light on the effect of board oversight on executive compensation. We find that firms that were affected the most by these requirements decreased the CEO compensation by about 25% after the regulations, compared to firms that were already complying with these requirements. The significant decrease in compensation is due to a decrease in the option-based portion of the compensation. The results suggest that board oversight is a significant determinant of the size and structure of executive compensation. The World Bank vchhaochharia@worldbank.org ** Cornell University, Johnson School of Management yg33@cornell.edu

Executive compensation has been a topic of considerable debate in recent years. The traditional view, which holds that compensation contracts to CEOs are determined mainly by supply and demand for executive talent and that incentive schemes are efficiently designed, was challenged recently by the view that supply and demand for executive talent has a limited effect on executive compensation, and that managers have a great influence over their own compensation, resulting in too-large compensation packages and inefficient incentive schemes (Bebchuk and Fried 2003, Bebchuk Fried and Walker 2002). According to this view, the reason managers have great influence over their own compensation is that the board of directors, who is supposed to be the representative of the shareholders and to structure the CEO compensation contracts efficiently, has little power and incentives to exercise its duty properly. There are several reasons for this lack of board oversight (Bebchuk and Fried 2003, Jensen 1993). One reason is that the nomination process in public U.S. firms gives CEOs a great influence over who will sit on the board, and directors who are on the board because of the CEO feel obligated to the CEO. Another reason is that board members are often very busy, and they have little time and incentives to negotiate with the CEO over compensation. A third reason is that shareholders have a hard time challenging compensation decisions or nomination decisions in the annual meeting, because such a challenge is often very costly. In this study we ask whether board oversight is an important determinant of executive compensation. The recent governance legislations of the U.S. exchanges offer a unique laboratory to test the effect of board oversight on CEO compensation. As a response to the corporate scandals, the Sarbanes Oxley law and the legislations of the major exchanges have put in place new rules, intended to enhance board effectiveness in 1

monitoring management. These changes include requirements such as changes to the director nomination process so that nomination of new directors is done by independent directors only, requirements for director independence on the compensation and audit committees, and a requirement for a majority of independent directors on the board. We hypothesize that if indeed lack of board oversight is the reason for too-large compensation arrangements then, holding else constant, we should expect firms that were the furthest away from complying with the regulations to decrease the CEO compensation after the regulations compared to other firms. By examining the compensation and board structure of about 900 firms, we find that firms that had to make the most changes to their board in order to comply with the rules (i.e., firms which did not have a majority of independent directors, an independent nominating committee, and independent compensation and audit committees), have drastically decreased the CEO compensation after the regulations compared to firms that were more compliant. The decrease is in the order of 20%-25%, after taking into account performance, size, industry effects, firm fixed effects, and other variables that affect compensation. In contrast, the compensation of firms that were more compliant did not decrease. Previous studies suggest that changes in the level of compensation do not capture the true effect on CEO utility, since compensation usually represent only a small portion of the managers company-related holdings (e.g., Core, Guay, and Verrecchia 2003). Since CEOs often hold a large amount of shares in the company, it makes sense to consider the change in CEO total wealth after the regulation. We find similar trends in executive wealth as we find in executive compensation. In the five years around the 2

regulation, CEO wealth in firms that had to make the largest changes to their boards drops by 25% more than the change in wealth of CEOs in firms that were already complying with the rules. We also look into the effect of board oversight on the different components of executive compensation. We find that the significant relative drop in the compensation comes from the decrease in the equity based portion of the compensation, particularly the decrease in option grants. This study belongs to a line of research that looks at the role of boards in influencing compensation contracts. Consistent with our results, studies by Hallock (1997) and Core et al. (1999) find that lack of independence of outside directors has a positive effect on executive compensation. Core et al. (1999), Cyert, Kang and Kumar (2002), and Grinstein and Hribar (2004) also find that CEOs who are chairmen of their boards receive higher compensation. All of the above studies which test the effect of board structure on CEO compensation face an identification challenge, since board structure is a variable partly determined by the CEO bargaining power, which in turn is partly determined by the CEO talent. For example, Hermalin and Weisbach (1998) provide a model where managerial talent increases managerial bargaining power over the filling of vacancies in the board of directors. Thus, firms in which managers have more talent will tend to have directors that are more linked to the CEO. But since talent is a variable that determines compensation, we should observe a relation between weak boards and compensation. Existing empirical studies try to control for this endogenous relation, but they have to rely on implicit assumptions about the relation between talent and measures of performance to control for 3

such endogenous relation. Unlike the above studies, our study looks the effect of an exogenous shock to board structure on executive compensation. Using a difference in differences approach we effectively control for any CEO specific attributes, such as talent, which have an effect on compensation. The rest of the study continues as follows. In the next section we describe the literature on board and compensation and the recent legislations. Section II describes the data and the variable. Section III has the results and Section IV has the robustness tests and Section V concludes. I. Board oversight and executive compensation review of the literature and recent regulations CEO compensation is a contract that is written between CEO of the firm and the board of directors, who are the representative of the shareholders. The traditional view is that CEO compensation is determined mainly by the supply and demand in the labor market for CEO talent. Therefore, a CEO whose skills are in short supply or who is required to exert higher effort is paid more for his or her services. Work by Mirrlees (1974, 1976), Holmstrom (1979), Grossman and Hart (1983), and others all show how to account for the moral hazard problem when designing the compensation contract. Since CEO s tasks are unobservable to the investors and the CEO prefers tasks that do not necessarily maximize investors wealth, firms should align managerial incentives by tying CEO compensation to observable outcome variables that are correlated with CEOs tasks. Compensation should therefore be based on observable measures of tasks that maximize value, such as market returns or profitability ratios. 4

In contrast to the above traditional view, a second view, the managerial power approach, argues that CEOs have the power to influence board decisions including compensation decisions, and that compensation contracts do not necessarily maximize shareholder wealth. According to this view, compensation arrangements are shaped both by market forces that push toward value-maximizing arrangements, and by the influence of managerial power over the board of directors, leading to departures from these arrangements in directions favorable to managers. The managerial power approach claims that these departures from value-maximizing arrangements are substantial (Bebchuk and Fried 2002). Bebchuk and Fried (2002) argue also that one of the main departures of compensation from value maximizations is the award of options that compensate the manager on absolute performance of the stock, rather than on the relative performance with respect to peers. Several studies have shown that measures of CEO power on the board of directors have a significant effect on CEO compensation. For example, Hallock (1997) looks at Forbes 500 firms in 1992 and finds that when the board has directors with interlocking relations the compensation to the CEO is higher. Core, Holthausen, and Larcker (1999), look at the level of compensation to CEOs in large U.S. firms in the mid 1980 s and they find that the level of CEO compensation is higher when the CEO has more power to affect board decisions. They use several measures to capture the CEO power in affecting board decisions, such as whether the CEO is involved in the nomination process of new directors, the percentage of affiliated directors on the board, whether the CEO is also the chairman of the board, and the number of directors on the board. They find that each of 5

the above variables is significantly positively related positively associated with the level of CEO compensation. Cyert, Kang, and Kumar (2002), look at the determinants of executive compensation in the early 1990 s in a large sample of U.S. public firms and they find that CEO who is also the chairman of the board receive higher compensation. Grinstein and Hribar (2004) look at the effect of CEO board power on the levels of the bonuses they receive for acquiring other firms. They find that the level of the bonus is higher when the CEO has more power to affect board decisions. The measures they find significant are whether the CEO is involved in the nomination process of new directors and whether the CEO is also the chairman of the board. In response to the Enron scandal, the Commission's Chairman at that time, Mr. Harvey Pitt, requested that the NYSE and NASD in February of 2002, as well as the other exchanges, review their governance listing standards.to further the ability of honest and well-intentioned directors, officers, and employees of listed issuers to perform their functions effectively. 1 In response, NYSE and NASDAQ came up with proposed changes in August 2002 (NYSE) and October 2002 (NASDAQ). These proposed changes were sent to the SEC for approval. The SEC approved these proposals with minor corrections in November 2003. The main provisions of the final NYSE rule are (Nasdaq and AMEX followed with similar rulings): 2 1. All firms must have a majority of independent directors. 1 Security and Exchange Commission, Release No. 34-48745. 2 NASDAQ relaxes some of the NYSE provisions to fit smaller firms. The main difference is that it also allows the compensation and nomination decisions to be held by a majority of independent directors without a formal committee, and it permits in special circumstances one non-independent board member to participate in these decisions. 6

2. Independent directors must comply with an elaborate definition of independence. 3. The compensation committee, nominating committee, and audit committee shall consist of independent directors. 4. All audit committee members should be financially literate. In addition, at least one member of the audit committee would be required to have accounting or related financial management expertise. 5. Separate executive sessions: The board should hold regular sessions without management. The aim of the above rules is to ensure efficient monitoring by corporate directors, and the rules should, arguably, reduce the influence that the manager has over the board of directors. For example, the requirements for an independent nominating committee and the requirement for a majority of independent directors on the board should reduce managerial influence over the nomination of new board members and should reduce the reliance of directors can feel for the CEO. The requirement for an independent compensation committee and the requirement for a committee charter should further remove any influence by the CEO or affiliated directors on compensation decisions. The NYSE has required firms to adopt these requirements until their first annual meeting after 1/15/2004 but not later than 10/31/2004. Firms with classified boards were given until the second annual meeting but not later than 12/31/2005 to comply with these requirements. We hypothesize that if board oversight has a large effect on CEO compensation, then the changes in board structure and director responsibilities in response to the new rules should have an effect on CEO compensation. To test this hypothesis, we construct a measure of the level of board compliance with the requirements before the rules were initiated. We use the difference-in-differences approach in our tests, where we test 7

whether firms which did not comply with the rules experience a significantly larger drop to their compensation after the rules relative to firms that did comply with the rules even before they were announced. We also check whether different components of the compensation are affected differently by the rules. To the extent that the option-based compensation provide suboptimal incentives, there should be a drop in that component of the compensation. The advantage of the difference-in-differences approach is that it controls for any firm specific or CEO specific unobservable characteristics that have an effect on CEO compensation. Since there are likely to be differences across firms in other characteristics besides board compliance, and these characteristics are likely to affect the level of compensation, it is important to control for them. Consistent with this argument, Murphy (1985) shows that it is important to control for firm fixed effects when analyzing CEO compensation to avoid wrong inferences. II. Data and Variables Our data source for executive compensation is the Execucomp database which has all compensation information about firms that belong to the S&P1500 index, or that once belonged to this index. Our data source for board structure and director information comes from the Investor Responsibility Research Center database (IRRC), which was recently bought by the Institutional Shareholder Services (ISS) company. The database includes information about directors in firms that belong to the S&P 1500 index. We have director information data for the years 2000 (before the rulings), 2003, and 2004 (after the rulings). To ensure that we do not capture changes in compensation due to firms 8

entering and leaving the samples, we include in the analysis only firms that existed in these two databases for the entire period. We retrieve financial information for each of the firms from Compustat. Our final sample has 896 firms. Our main variable of interest is the CEO compensation. We use the total compensation of the CEO, which includes their base salary, bonuses, options (Black Scholes value), restricted stocks, and other compensation. This variable is referred to as TDC1 in the Execucomp database. We also look separately at the equity based compensation, and at the cash-based compensation. The equity based compensation is defined as the total value of options (Black Scholes value) and restricted stock to the CEO, and the cash based compensation is the total compensation to the CEO minus the equity based compensation. In our analysis, we also look at the total wealth of the CEO that is tied to the firm. To calculate CEO wealth as the total value of the firm s stock and options that the CEO owns at the end of the year plus any cash-based compensation that the CEO received during the year plus any cash inflow or outflow from redemptions of options and stock during the year. We use the approximation procedure offered by Core and Guay (2002) to calculate the value of the non-vested and vested CEO options that the CEO holds at the end of the year. (For a description of the methodology se the appendix). We define the change of CEO wealth between date1 and date2, as the CEO wealth in date2 plus any cash compensation that the CEO receive between date 1 and date 2 plus any cash inflow or outflow from redemptions of options and stock between date1 and date2 minus CEO wealth in date 1. 9

To measure the level of board oversight we use an index that captures the compliance of the firm with the independence requirements of the board. The index is the sum of four indicator variables for whether the firm had in the year 2000 (before the rulings) an independent nominating committee, an independent compensation committee, an independent audit committee, and a majority of independent directors on the board of directors. This measure was used by Chhaochharia and Grinstein (2005), to measure level of compliance with the rules. The drawback of an index, is that it assigns ad-hoc weights to the different provisions. To the extent that certain provisions of the rules are more important than others in affecting board oversight, the index will not capture the true level of oversight. For example, if majority of independent directors is the most important provision, then firms that comply only with this provision might have stronger board oversight then firms that comply instead with the requirements for independent audit committee and compensation committee even though their score will be lower. To avoid such possibility, and to increase the power of our tests, we define non-complying firms as firms that do not comply with any of the provisions. Regardless of the relative importance of the different provisions, these firms will have lower board oversight than other firms since they do not comply with any of the provisions. We also use control variables in the different tests. To control for firm size we use the sales of the corporation (in $millions). To control for performance we use the annual stock return (dividends reinvested) of the firm, ending in the beginning of the fiscal year. We also use the return on assets, which is defined as the net income before extraordinary items divided by the book value of assets. To control for CEO tenure, we use a dummy 10

variable NEW CEO that equals 1 if the CEO served less than 2 years. Finally, we include industry dummies, using the Fama and French (1997) 48 industry classification. Table 1 shows summary statistics of firms in our sample for each of the years 2000-2004. Panel A shows the financial characteristics of the firms. Average sales are around $6 billion, increasing to about $7 billions in 2004. Median sales are much lower (in the order of $1.7 billion). The difference between the average and median sales suggest that the sample is skewed by several very large firms. Consistent with the downturn in the economy between 2000-2002 and the upturn between 2003-2004, market value has decreased between 2000-2002 and then increased between 2003-2004. Returns on assets and stock returns show also a similar pattern. Table 1 panel B shows information about the compensation of CEOs in our sample. Average total compensation is $9.6 million in 2000 and then drops to $6.3 million in 2002 and to $5.5 million in 2003. The average compensation increases to $6.18 million in 2004. Median compensation, however almost does not change over the years and is between $3 million and $4 million throughout the years. Average equity-based compensation has decreased significantly over the years from $7.3 millions in 2000 to $3.4 million in 2004. Consistent with this trend, the average portion of equity-based compensation has decreased from 51% in 2000 to 46% in 2004. Panel C shows governance characteristics of firms in the sample. We see a significant increasing trend in the % of firms in the sample that have independent nominating, compensation, and audit committee, and in the percentage of firms that have a majority of independent directors. Between 2000 and 2004, the percentage of firms with independent nominating committee has increased from 28% to 73%, the percentage 11

of firms with independent compensation committee has increased from 71% to 82%, the percentage of firms with independent audit committee has increased from 63% to 83%, and the percentage of firms with a majority of independent directors has increased from 73% to 88%. All of these increases are statistically significant at the 1% level. The average score has also increased from 2.36 in 2000 to 3.26 in 2004, and the percentage of firms with a score of 0 (not complying with any of the requirements) has decreased from 12% to 4% between 2000 and 2004. Average board size has also decreased in recent years from 9.74 to 9.36, and the percentage of firms in which the CEO is a member of the nominating committee or that there is no nominating committee increased significantly from 52% to 3%. Unlike the significant trends in board and committee independence, there is no significant change in the percentage of firms that have a chairman CEO. In 2004, 64% of the firms in the sample have a CEO who is also the chairman. III. Results A. Changes to total compensation We use several specification for our tests. The first specification is a fixed effect regression as shown is equation (1). COMP it =a 0 + a 1 SALES it + a 2 ROA it + a 3 RET it + [Year Dummies] + (1) a 4 Dummy2003_2004*Score0 i + a 5 NEW CEO it + v i + ε it 12

Where COMP is the natural log of CEO total compensation in year t, SALES is the natural log of total sales, ROA is the natural log of one plus the return on assets, and RET is the natural log of the gross stock return. Dummy2003_2004*Score0 i is an interaction dummy that is composed of Dummy2003_2004, which equals 1 if the year is either 2003 or 2004 and zero otherwise, and Score0, which equals 1 if the firm did not comply with any of the requirements (score=0) in the year 2000 (before the rules). NEW CEO is a dummy variable that equals 1 if the CEO tenure in the firm is less than two years. The variable v i is a firm-level fixed effect. If indeed board oversight is going to affect compensation, then firms that did not comply with the rules are going to decrease their compensation after the rules (compared to their average level of compensation) more than other firms. This prediction would mean that we should expect the a 4 coefficient to be significantly positive. The rules have been put in place already in 2003, but firms were given a couple of years to comply with the rules. If indeed board oversight affects compensation, we should expect an effect on compensation after the rules only in firms that changed their governance structure after the rules. To test this hypothesis we decompose the interaction term Dummy2003_2004*Score0 i in regression (1) into Dummy2003_2004*Score0_Changed and Dummy2003_2004*Score0_Changed_Not_Changed. The variable Score0_Changed is a dummy variable that equals 1 if the firm had a score of 0 in the year 2000 but has improved since then and zero otherwise. The variable Score0_Changed is a dummy variable that equals 1 if the firm had a score of 0 in the year 2000 but has not improved by 2003. 13

Table II column 1 shows the results of regression (1), and Table II panel 2 shows the results of the decomposition of the interaction dummy. As expected, in both cases, size and performance have a significant positive effect on compensation. The dummy for the year 2000 is also significantly positive, suggesting that compensation in the year 2000 is also significantly higher than in the rest of the years. The interaction dummy is significantly negative (10% significance level), with a coefficient of -0.092. This result suggests that firms with the lowest board score had decreased the compensation to their CEO by more than other firms. Since this is a log-log regression, the coefficient suggests that firms that did not comply with the rules had a drop in their compensation that is 9.2% larger compared to other firms. Table II column 2 shows that when decomposed, the only the interaction dummy of firms that changed their governance structure is significantly negative. Firms that did not change their governance structure did not experience a decrease in compensation compared to other firms. The sample used for the regression includes all firms that existed between 2000 and 2004. These firms include also firms that had replaced their CEO during that time. Thus, it could be that the NEW CEO dummy does not fully capture the effect of new CEOs on compensation and instead we capture a change-in-ceo effect in the interaction dummy. We therefore run also the regression above on a sample of firms that did not replace their CEOs. This requirement reduces the sample size from 896 to 366. We show the regression results on this sample in Table II columns 3 and 4. The results are much stronger when we consider only firms in which the CEO has not been replaced. The coefficient of the interaction dummy is -0.209 and is significant at 14

the 1% level. When we decompose the interaction dummy we find that all of the effect is coming from the firms that changed their governance structure. The coefficient there is - 0.249 which suggests a relative drop in the order of 25%. An alternative specification is the following one COMP i2003-2004 - COMP i2000-2001 =a 0 + a 1 SALES i2000-2001 + a 2 ROA i2000-2001 (1) + a 3 RET i2000-2001 + a 4 (SALES i2003-2004 - SALES i2000-2001 )+ a 5 (ROA i2003-2004 - ROA i2000-2001 )+ a 6 (RET i2003-2004 - RET i2000-2001 )+a 7 Score0 i + a 8 NEW CEO i + [INDUSTRY DUMMIES] + ε i The variable COMP i2000-2001 is the natural log of the average compensation to the CEO over the years 2000-2001, SALES i2000-2001 is the natural log of the average sales over these years, RET i2000-2001 is one plus the average gross return over these years and ROA i2000-2001 is the natural log of one plus the average return on assets over these years. For variables with a subscript 2003-2004, the averaging is over the years 2003-2004. If indeed board oversight is going to affect compensation, then firms that did not comply with the rules are going to decrease their compensation after the rules (compared to their average level of compensation) more than other firms. This prediction would mean that we should expect the a 7 coefficient to be significantly negative. Table III shows the results of regression (2) on the sample of 366 firms that did not replace their CEOs throughout the years. Column 1 shows that the coefficient of the dummy variable Score0 is -0.18 and is significantly negative at the 5% significance level. This result means that firms that were not complying with the rules have decreased their compensation by 18% more than other firms. Table III column 2 shows the results where 15

the score dummy is decomposed to a dummy of firms that changed their governance structure and firms that did not change the structure. All of the effect is concentrated in firms that changed their board structure. The coefficient is -0.203 and is significantly negative at the 5% significance level. One of the criticisms over the specifications in regressions (1) and (2) is that they do not capture the real effect of board oversight on CEO wealth. Theory suggests that boards should be concerned with the effect of their incentive compensation on total CEO wealth (or utiliy). Since much of the CEO wealth is in the form of stock and non-vested options holdings compensation, and annual compensation is only a small part of CEO wealth, it makes sense to look at the total effect of board oversight on CEO wealth. We therefore run the following regression specification WEALTH i2000_2004 =a 0 + a 1 SALES i2000 + a 2 ROA i2000 + a 3 RET i2000 + (3) a 4 (SALES i2004 - SALES i2000 )+ a 5 ROA i2000_2004 + a 6 RET i2000_2004 +a 7 Score0 i + a 8 NEW CEO i + [INDUSTRY DUMMIES] + ε i Where WEALTH i2000_2004 is the natural log of the value of CEO stock holdings and options in the year 2004 plus all cash inflows and outflows from cash compensation and redemptions of options and stock between 2000-2004 minus the log of the value of CEO stock holdings and options in the year 2000 minus any cash from redemption of options or from cash compensation in the year 2000. ROA i2000_2004 is the natural log of one plus the cumulative return on assets between 2000-2004, and RET i2000_2004 is the natural log of one plus the cumulative gross stock return between 2000-2004. 16

The results of this specification appear in table IV column (1). The results suggest that growth in sales and cumulative returns have a positive effect on the CEO wealth. However, for firms that were the least compliant, there was a significant decrease in the compensation. The coefficient of the dummy Score0 is -0.23 and is significant at the 10% level. Column 2 shows the results where we decompose the score into firms that changed and firms that did not change their board structure. We find that all of the effect is concentrated in firms that changed their board structure. The coefficient is -0.243 and is significantly negative at the 5% level. Overall, the results in tables II-IV suggest that the total compensation to CEOs of firms that were the least compliant and changed their governance structure after the rules has decreased by between 20%-25%, after controlling for the decrease in compensation to all other firms. B. Changes to cash based compensation and equity based compensation Bebchuk and Fried (2003), argue that powerful CEOs are likely to manipulate their compensation schemes in ways that are least transparent, and which lead to the least concerns by shareholders. This argument implies that powerful CEOs are more likely to award themselves what looks like incentive compensation because such compensation is less likely to cause concerns by shareholders. If this argument is valid, then we should observe a decrease in the equity-based portion of the compensation rather than in the cash-based portion. 17

To test which component of the compensation is affected, we repeat the analysis in section I, but this time we run separately the regressions once with the cash-based portion of the compensation as the dependent variable, and once with the equity-based portion of the compensation as the dependent variable. Table V shows the results where the dependent variable is the cash-based component of the compensation. Panel A shows the results of the fixed-effect regression. Since previous results suggest that only the firms that changed their structure decreased the compensation, we run only the regression with the decomposition of the interaction dummy. The results show no significant drop in the cash portion of the compensation both in the entire sample (column 1) and in the sample of CEOs that were not replaced (column 2). The results are similar when we run the change-in-compensation regression (panel B). The score0 coefficient is insignificant. Table VI shows the results where the dependent variable is the equity-based component of the compensation. The fixed effect regression shows a significantly negative coefficient of the interaction dummy of firms that changed their structure. The coefficient is -0.564 in the sample of all firms, and -0.65 in the sample of firms that did not replace their CEOs. Both coefficients are significant at the 5% level. The magnitude of the coefficient suggests that the decrease in equity-based compensation was in the order of 56% - 65%. We obtain similar results when we consider the change in compensation regression. Panel B shows that the coefficient of the. The results show a significantly negative coefficient of the interaction dummy of firms that changed their board structure. 18

C. Changes to the option based and the stock based compensation Since most of the change in compensation occurred in the equity based compensation, it is natural to explore which of the components of the stock based compensation has decreased. We therefore decompose the equity based compensation into option based and stock based compensation, and use each one separately as the dependent variable. Table VII shows the results of the option based compensation. The fixed effect regression in panel A shows a significant effect (at the 10% level) on Dummy03_04*Score0_Changed when all firms are considered (coefficient of -0.472), and negative and insignificant effect when only the firms that were not replaced are considered. The change-in-compensation regressions also shows a significant negative effect (at the 10% level). Table VIII shows the results of the stock-based compensation. The fixed-effect regression (panel A) shows that the coefficient of the interaction dummy are negative but insignificant in either case. The change in compensation regression (panel B) shows also negative but insignificant effect on the change in compensation. Overall the results suggest that the driver for the drop in the compensation is the drop in the option-based compensation. However, the results are only marginally significant, and the negative effect appears also in the stock based component in the compensation. IV. Robustness check - Matched Sample methodology The regression specifications (1) and (2) assume a log-log relation between compensation and size. They also assume that industry has a constant effect on compensation regardless of the year. As a robustness check, we use a matched sample methodology, where we 19

match firms that had a [L]ow score (0,1) in the year 2000 to firms that had a [H]igh score (3,4) in the year 2000 based on industry and size. We match first on industry and then choose the firms that are closest in size. We include only firms that did not replace their CEOs. A total of 82 firms are in each matched sample. We then compute the average compensation in each of the samples for each of the years. We plot the results in Figure 1. The figure shows the moving averages of the average compensation of each group. The figure shows that in the year 2000 firms that belonged to group L had average compensation than was about 25% higher then the compensation of CEOs in firms that belonged to group H. However, the gap has decreased and by 2003-2004 the average compensation was almost the same. The difference in the levels of the compensation from 2000-2001 to 2003-2001 in the group of firms that were less compliant is significant at the 1% level. There is no difference in the levels of the compensation for the firms that were more compliant. Figure 2 and Figure 3 show differences in size and in performance across the two groups over the same period. There are no significant differences in the average performance across the groups over the years, suggesting that the effect is unlikely to come from differences in performance of changes in sizes over the years. V. Conclusion The new requirements of the exchanges from boards of directors led to significant changes in board structure in U.S. corporations. Using the difference in differences approach we find that firms that were least compliant with the rules decreased the compensation to the CEO once they started to comply with the rules. The decrease was in 20

the order of 20%-25% over and above the decrease in firms that were more compliant before the rules. The decrease was mainly in the form of cutting the options to the CEOs. These results corroborate the arguments by Bebchuk and Fried that board oversight has a significant effect on the size and structure of executive compensation. The results also corroborates the study of Chhaochharia and Grinstein (2006) which shows that firms which did not comply with the exchange requirements, enjoyed a positive abnormal return upon the announcement of these rules. One potential source of this increase in value is the reduced compensation to the CEO, and potentially better compensation schemes. 21

Appendix: Procedure of calculating changes in CEO wealth In calculating the value of the current CEO holdings of the company, we follow closely Core and Guay (2002). The current holdings of the CEO in any given year consists of his restricted stock and his options outstanding. The value of the restricted stock is calculated using the share price at the end of the fiscal year, multiplied by the number of stocks outstanding. To evaluate the options we use the Black-Scholes (1973) model, as modified by Merton (1973), to account for dividend payouts. Following Core and Guay (2002), we calculate separately the value of the new option grants that were given during the year and the option grants that were given in previous years. To calculate the value of new option grants during the fiscal years we use the information in Execucomp. Execucomp provides all the details necessary to calculate the value of stock option grants that were given during the year, including the strike price [EXPRIC], the expiration date [EXDATE], the expected standard deviation of the stock price [BS_VOLAT] and the expected dividend yield [BS_YLD], and the fiscal-year end stock price [PRCCF]. The risk-free rate is the treasury yield corresponding to the time-tomaturity of new options, which is readily available from CRSP. For previously granted options, Execucomp (and the proxy statements), do not disclose the strike prices and the maturity time of the options. To actually find out the strike and maturity of the options, one would need to follow the option grants and the insider selling statements of each CEO over the years to subtract one from the other. This task is tedious and might not lead to accurate results (due to inaccuracy in some of the insider selling statements). To overcome this problem, Core and Guay (2002) offer an 22

approximation method. The approximation method uses the information available in the proxy statements about previously granted options: a. Number of options that are exercisable. b. Number of options that are not exercisable. c. Total value of all exercisable options if they were realized on the last day of the fiscal years. (This of-course includes the value of all in-the-money options, since the realization value of out-of-the-money options is zero). d. Total value of all non-exercisable options if they were realized on the last day of the fiscal year. To approximate the average strike price of the previously-granted exercisable options, divide the total value of exercisable options by the number of exercisable options and add the stock price at the end of the fiscal year. To approximate the average strike price of the previously-granted non-exercisable options, divide the total value of non-exercisable options by the number of non-exercisable options and add the price at the end of the fiscal year. In effect, this approximation method assumes that the strike price of all outof-the-money options have a strike price that equals the stock price at the end of the fiscal year. Core and Guay compare this approximation method to an actual hand-collected sample of previously issued options and find little bias and very high correlation (explained variation of 99%). To approximate the time to maturity of previously-granted non-exercisable options, use the time-to-maturity of new options and subtract one year. If no options are granted in the current year, use 9 years. 23

To approximate the time-to-maturity of previously-granted exercisable options, use the time to maturity of the previously-granted non-exercisable options and subtract three years. 24

References Bebchuk, L.A., Fried, J.M., 2003. Executive compensation as an agency problem. Journal of Economic Perspectives 17:3, 71-92. Bebchuk, L.A., Fried, J.M., Walker, D.I., 2002. Managerial Power and Rent Extraction in the Design of Executive Compensation. The University of Chicago Law Review 69, 751-846. Blanchard, O.J., Lopez De Silanes, F., Shleifer, A., 1994. What do firms do with cash windfalls? Journal of Financial Economics 36, 337-60. Chhaochharia, V., Grinstein Y., 2006, Corporate governance and firm value the impact of the 2002 governance rules, Journal of Finance, forthcoming. Core, J.E., and Guay, W., 2002, Estimating the value of employee option portfolio and their sensitivity to price and volatility, Working paper, University of Pennsylvania. Core, J.E., Guay, W., and Verrecchia, R.E., 2003, Price versus Non-Price Performance Measures in Optimal CEO Compensation Contracts, The Accounting Review 78, 957-981. Core, J.E., Holthausen R.W., and Larcker D.F., 1999. Corporate governance, CEO compensation, and firm performance. Journal of Financial Economics 51, 371-406. Cyert, R., Kang, S., and Kumar, P., 2002. Corporate governance, takeovers, and topmanagement compensation: Theory and evidence. Management Science 48, 453-469. Grinstein, Y., and Hribar, P., 2004 CEO compensation and incentives: evidences from M&A bonuses, Journal of Financial Economics, 73, 119-143. Grossman, S., Hart, O., 1983. An analysis of the principal agent problem. Econometrica 51, 7-45. Hallock, K. F., 1997, Reciprocally interlocking boards of directors and executive compensation, Journal of Financial and Quantitative Analysis, 32, 331-344. Hermalin, B.H., Weisbach, M.S., 1998. Endogenously chosen boards of directors and their monitoring of the CEO. American Economic Review 96. Holmstrom, B., 1979. Moral hazard and observability, Bell Journal of Economics 13, 234-340. 25

Jensen, M.C., 1993. The modern industrial revolution, exit, and the failure of internal control systems. Journal of Finance 48, 831-880. Mirrlees, J., 1974. Notes on welfare economics, information, and uncertainty, in Balch, M., McFadden, D., Wu, S., eds, Essays on Economic Behavior Under Uncertainty. North Holland, Amsterdam. Mirrlees, J., 1976. The optimal structure of incentives and authority within an organization. Bell Journal of Economics 7, 105-131. Shivdasani, A., Yermack, D., 1999. CEO involvement in the selection of new board members: An empirical analysis. Journal of Finance 54, 1829-53. Yermack, D., 1995. Do corporations award CEO stock options effectively? Journal of Financial Economics 39, 237-269. Yermack, D., 1996. Higher market valuation of companies with a small board of directors, Journal of Financial Economics 40, 185-211. Yermack, D., 1997. Good timing: CEO stock option awards and company news announcements. Journal of Finance 52, 449-476. 26

Table I: Summary Statistics The table shows financial, compensation, and governance characteristics of U.S. public firms between 2000 and 2004. The sample consists of 896 firms which have executive compensation information as well as board structure information throughout the years 2000-2004. The numbers without parentheses are averages, and the numbers within parentheses are medians. In Panel A, market value is the market capitalization of equity. Return on assets is the the net income before extraordinary items and discontinued operations divided by the book value of assets, and stock return is the annual stock return (dividend reinvested). Total In panel B, Compensation is variable TDC1 in Execucomp, which consists of salary, bonus, value of restricted stock granted, value of options granted (using Black Scholes), long term incentive payouts and other compensation. Total Equity-based Compensation is the value of restricted stock and options granted. % Equity-based Compensation is the Equity-based Compensation as a percentage of Total Compensation. In panel C, the board score is the sum of four indicator variables: existence of a majority of independent directors, existence of an independent audit committee; existence of an independent nominating committee and existence of an independent compensation committee. The last column shows the significance of a non-parametric binomial test of difference in probabilities. *** indicates significance at the 1% significance level. Panel A: Financial Characteristics. Year 2000 2001 2002 2003 2004 Sales ($million) 6248 6337 6080 6379 7166 (1764) (1773) (1698) (1850) (2055) Market Value ($million) 12075 10318 7977 9684 10769 (2175) (2284) (1771) (2459) (2779) Assets ($million) 13928 15446 16209 17241 20544 (2019) (2248) (2384) (2608) (2932) Return on Assets (%) 6.1 2.9 3.1 4.0 5.1 (5.2) (3.6) (3.9) (4.1) (4.9) Stock Return (%) 23.1 11.1-9.8 39.6 19.3 (12.5) (5.9) -(8.2) (30.7) (16.8) Panel B: CEO Compensation. Year 2000 2001 2002 2003 2004 Total Compensation 9638 7937 6318 5522 6180 (3444) (3546) (3497) (3265) (4043) Total Equity-based Comp. 7375 5706 4141 3120 3459 (1677) (1840) (1725) (1389) (1755) % Equity-based Comp. 0.51 0.53 0.50 0.45 0.46 (.52) (.59) (.53) (.48) (.49) 27

Panel C: Governance Characteristics Year Difference 2000 2003 2004 2004-2000 Independent nominating committee 28% 57% 73% *** Independent compensation committee 71% 77% 82% *** Independent audit committee 63% 76% 83% *** Majority of independent directors 73% 83% 88% *** Score (sum of the four variables) 2.36 2.93 3.26 *** % with Score=0 12% 7% 4% *** CEO Chairman 67% 66% 64% CEO is on the nom. Committee / no nom. Committee 52% 14% 3% *** Board size 9.74 9.37 9.36 *** 28

Table II: CEO Compensation and Board Compliance - Fixed Effect Regression The table shows the results of a fixed-effect regression, where the dependent variable is the natural log of total CEO compensation. Columns 1 and 2 show the results of the entire sample of 896 firms that existed in the IRRC and Execucomp databases throughout the period 2000-2004. Columns 3 and 4 show the results of a subsample of 366 firms where the CEO has not been replaced during the period 2000-2004. SALES is the natural log of the company sales (Compustat data item 12), ROA is the natural log of one plus the net income before extraordinary items and discontinued operations divided by the book value of assets - all measured in the beginning of the fiscal year. RET is the natural log of the annual gross stock return (dividend reinvested), measured in the beginning of the fiscal year. Dummy NEW CEO equals 1 if the CEO tenure is less than two years. Score0 is a dummy variable that equals 1 if the firm had a governance score of 0 in the year 2000 and 0 otherwise. Score0_Changed equals 1 if the firm had a score of 0 in 2000 and had increased its score by 2003. The Score is a sum of four dummy variables of board and committee independence, and it includes a majority of independent directors on the board, an independent compensation committee, an independent nominating committee, and an independent audit committee. Dummy 03_04 is a dummy variable that equals 1 if the current year is 2003 or 2004 and 0 otherwise. *,**,*** indicates significance at the 10%, 5%, and 1% levels respectively. Variable (1) (2) (3) (4) SALES 0.353 *** 0.352 *** 0.420 *** 0.419 *** (0.046) (0.046) (0.059) (0.059) ROA 0.343 *** 0.343 *** 0.349 0.339 (0.109) (0.109) (0.233) (0.233) RET 0.166 *** 0.166 *** 0.181 *** 0.181 *** (0.021) (0.021) (0.028) (0.028) Dummy year 2000 0.058 ** 0.058 ** -0.046-0.047 (0.028) (0.028) (0.036) (0.036) Dummy year 2001 0.026 0.026-0.062 * -0.062 * (0.027) (0.027) (0.035) (0.035) Dummy year 2002 0.011 0.010-0.059 * -0.059 * (0.027) (0.027) (0.035) (0.035) Dummy year 2003-0.015-0.015 0.001 0.001 (0.027) (0.027) (0.034) (0.034) Dummy 03_04*Score0-0.092 * -0.209 *** (0.054) (0.065) Dummy 03_04*Score0_Changed -0.118 * -0.249 *** (0.061) (0.072) Dummy 03_04*Score0_Not Changed -0.010-0.068 (0.105) (0.131) Dummy NEW CEO 0.030 0.029-0.078 * -0.077 (0.025) (0.025) (0.046) (0.046) 29

Table III: Changes in CEO Compensation and Board Compliance Regression Results The table shows the results of an ordinary least-square regression, where the dependent variable is the log(average total CEO compensation in the years 2003-2004) log(average total CEO compensation in the years 2000-2001). Column 1 shows the results of the entire sample of 896 firms that existed in the IRRC and Execucomp databases throughout the period 2000-2004. Column 2 shows the results of a subsample of 366 firms where the CEO has not been replaced during the period 2000-2004. SALES 2000-2001 is the natural log of the average company sales over the years 2000-2001 (Compustat data item 12). ROA 2000-2001 is the natural log of one plus the average ROA of the firm over the years 2000-2001 where ROA is the net income before extraordinary items and discontinued operations divided by the book value of assets - all measured in the beginning of the fiscal year. RET 2000-2001 is the natural log of the average annual gross stock return (dividend reinvested), over the two years that end in the beginning of fiscal year 2001. All the above variables with a subscript of 2003-2004 are averages over the years 2003-2004. Dummy NEW CEO equals 1 if the CEO tenure is less than two years. Score0 is a dummy variable that equals 1 if the firm had a governance score of 0 in the year 2000 and 0 otherwise. Score0_Changed equals 1 if the firm had a score of 0 in 2000 and had increased its score by 2003. Score0_Not Changed equal1 1 if the firm had a score of 0 both in 2000 and in 2003. The Score is a sum of four dummy variables of board and committee independence, and it includes a majority of independent directors on the board, an independent compensation committee, an independent nominating committee, and an independent audit committee. Dummy 03_04 is a dummy variable that equals 1 if the current year is 2003 or 2004 and 0 otherwise. *,**,*** indicates significance at the 10%, 5%, and 1% levels respectively. (1) (2) Intercept 0.302 0.290 (0.247) (0.247) SALES 2000-2001 -0.010-0.008 (0.022) (0.022) RET 2000-2001 0.539 ** 0.542 ** (0.224) (0.224) ROA 2000-2001 0.077 0.026 (0.56) (0.564) SALES 2003-2004 - SALES 2000-2001 0.393 *** 0.396 *** (0.104) (0.104) RET 2003-2004 - RET 2000-2001 0.652 *** 0.650 *** (0.183) (0.183) ROA 2003-2004 - ROA 2000-2001 0.479 0.442 (0.675) (0.677) Dummy Score0-0.180 ** (0.088) Dummy Score0_Changed -0.203 ** (0.092) Dummy Score0_Not Changed 0.034 (0.263) Dummy NEW CEO 0.082 0.081 (0.065) (0.065) Industry Dummies (48) + + R Square 23% 23% 30