Board meetings, committee structure, and firm value

Board meetings, committee structure, and firm value

Journal of Corporate Finance 16 (2010) 533–553 Contents lists available at ScienceDirect Journal of Corporate Finance j o u r n a l h o m e p a g e ...

335KB Sizes 0 Downloads 64 Views

Journal of Corporate Finance 16 (2010) 533–553

Contents lists available at ScienceDirect

Journal of Corporate Finance j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j c o r p f i n

Board meetings, committee structure, and firm value Ivan E. Brick a,⁎, N.K. Chidambaran b,1 a b

Rutgers Business School - Newark & New Brunswick, Rutgers University, Newark, NJ 07102, United States Graduate School of Business, Fordham University, New York, NY 10023, United States

a r t i c l e

i n f o

Article history: Received 9 September 2008 Received in revised form 7 June 2010 Accepted 8 June 2010 Available online 12 June 2010 JEL classification: G34

Keywords: Corporate governance Board meetings Board committees Firm value

a b s t r a c t In this study, we examine the determinants of board monitoring activity and its impact on firm value for a broad panel of firms over a six-year period from 1999 to 2005. During this period, Congress and the exchanges promulgated regulations that increased pressure upon firms for more independent and active boards. Economists have debated whether board activity and externally imposed regulations benefit or harm firms. We develop and examine several proxies for board monitoring and examine the relationship between board monitoring activity, firm characteristics, and firm value in a structural equation framework. One set of our proxies is based on the number of annual board and Audit Committee meetings. We show that prior performance, firm characteristics and governance characteristics are important determinants of board activity. We also show that the board monitoring is driven by corporate events, such as an acquisition or a restatement of financial statements. We find that board activity has a positive impact on firm value. Our results also indicate that the external pressure has had a salutary effect and recent regulations have led to some increase in firm value. A second set of proxies is based on the shift to a fully independent Audit, Compensation and Nominating Committees. We find that firms increased the independence of these Board committees following the enactment of the 2002 Sarbanes-Oxley Act. © 2010 Elsevier B.V. All rights reserved.

1. Introduction Oversight by the board of directors has received increasing attention by the business press and community beginning in the late 1990s. Board monitoring has also been a popular topic in the academic literature. The literature has mostly focused on the size and composition of the board as measures of board involvement in the firm (see Linck et al., 2008). An additional, perhaps equally important dimension of board oversight is the intensity of board activity, which encompasses the frequency of board meetings and the changes in structure of board subcommittees. Such board activity has received intense scrutiny by regulators and shareholders and firms now routinely report on the number of annual meetings and composition of the boards committees in the proxy statements filed with the SEC (Form DEF 14A). Governance reformists and shareholder services groups such as the RiskMetrics Group and The Corporate Library, collate such data in order to evaluate the effectiveness of the board of directors and make the argument that board activity is very important and material in valuing the firm. In this paper, we empirically examine the determinants of board activity and the effect of board activity upon firm value. Our analysis is over the seven-year period from 1999 to 2005, a time period that has seen the increased attention upon the role of the board.2

⁎ Corresponding author. Tel.: + 1 973 353 5155; fax: + 1 973 353 1006. E-mail addresses: [email protected] (I.E. Brick), [email protected] (N.K. Chidambaran). 1 Tel.: + 1 646 312 8248; fax: + 1 646 312 8245. 2 In 1998, the NYSE Blue Ribbon Commission recommended that the Audit Committee be composed of only independent directors. The 2002 Sarbanes-Oxley Act requires that all Audit Committee members be independent, financially literate and that at least one member have accounting or financial management expertise. The law also requires the CEO and the board to sign off on financial statements. In 2004, the NYSE started to require its listed companies to disclose as to whether the firm had a separate nominating committee. 0929-1199/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.jcorpfin.2010.06.003

534

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

Our focus on the determinants of board activity complements and extends the literature that examines the determinants of board meetings. Poorly performing firms could be obvious candidates for greater board monitoring, as found by Vafeas (1999) and Adams (2005). It is also plausible that litigation risk increases when the firm does poorly; we therefore expect that boards will increase their activity to insulate themselves from the charge that they were asleep at the wheel (see Brick and Chidambaran, 2008). We expand on the variables used in the literature for determining the level of board monitoring, by including variables that proxy for the increased need for information. We would expect the board to meet more if firms are involved in a major investment program such as in a merger or acquisition or if the firm is forced to restate earnings. Similarly, since the board is responsible to provide strategic advice to management, we would expect board activity to increase as investment opportunities increase. Much of the regulatory and shareholder attention on the board has been predicated on the assumption that board activity can enhance shareholder value. We empirically examine this issue and ask whether board monitoring increases firm value. Our data sample is also useful in answering an important ancillary question as our sample period straddles the passage of the 2002 Sarbanes-Oxley Act. Was the increased political and regulatory scrutiny of board activity that especially followed the passage of Sarbanes-Oxley beneficial to firm value? Regulatory reform can enhance firm value as it has enhanced the monitoring and advisory role of the board, which can reduce agency costs. On the other hand externally imposed regulation on board activity can be costly and can have unintended consequences, as Hermalin and Weisbach (2006) argue. More independent boards can also have higher monitoring costs because it is more costly to gather and communicate relevant information to outside board members (see Raheja, 2005). Furthermore, if the impetus behind board activity were simply the need to comply with regulation and the fear of stockholder litigation, increases in board activity could have a negative impact on firm value as the increased activity would detract management from focusing on running the firm. We use two sets of proxies to measure board activity. Our first set of proxies is related to the number of annual board and committee meetings as in Vafeas (1999). We use the log of the number of annual board meetings and the log of the number of “director-days”, which is the product of the number of meetings and the number of independent directors. It is plausible that the number of meetings alone does not fully capture the level of board activity, and that both the number of independent directors and the time they spend on monitoring are important. Since a major element of the Sarbanes-Oxley Act is the increased emphasis on Audit Committees, we also examine the monitoring role of the Audit Committee using proxies similar to our board proxies. Our second set of proxies uses the changes in the board's committee structure. Using data on committee size and composition from company filings, and the classification of directors as outsiders by the RiskMetrics Group, we document the firm's move to fully independent Audit, Compensation and Nominating committees. Our proxies are motivated by literature that has shown that committee independence may represent increased monitoring by the board and can affect value. Klein (2002) finds that earnings management decreases with the independence of the Audit Committee. Jensen (1993) argues that the board effectiveness can be compromised if the CEO controls the composition of the board. Chhaochharia and Grinstein (2007) show that firms who are not in compliance with respect to the committee independence as required by the enactment of the 2002 Sarbanes-Oxley Act have superior returns following the public announcement of the Act's passage. Our work extends their work by examining the empirical determinants of the firm's decision to move to fully independent committees in the years prior to and following the passage of the Act and the impact of the move on firm value. We use the industry-adjusted Tobin Q (at the 2-digit SIC level), where the firm's Tobin Q is the ratio of the market value of the firm to the book-value of total assets, as our measure of firm value. Tobin Q has been frequently used in the corporate governance literature (see Yermack, 1996) to proxy firm value. Our empirical methodology takes into consideration that the relationship between board activity and firm value is endogenous. The analysis of board monitoring also cannot be independent of incentive compensation contracts that align the CEO and shareholders. We follow a structural model approach in our analysis with instrumental variables that are motivated theoretically and empirically by the literature. As in Vafeas (1999) and Adams (2005), we find that the level of board monitoring is inversely related to the firm's prior performance. Poorly performing firms increase the board's monitoring activity. In addition to these findings, we also report several new results. We find that the level of board monitoring activity is driven by merger and acquisition activity and by accounting restatements. We also find that board activity increased significantly following the passage of Sarbanes-Oxley. These findings are consistent with the notion that boards facing increased political pressure increase their monitoring. The increases in board activity for the full Board result in increases in firm value after controlling for the endogeneity between firm performance, monitoring and CEO pay-performance sensitivity. The increased monitoring following the passage of the Sarbanes-Oxley also had a positive impact. Our findings are thus consistent with external pressure having a salutary impact on firm value. As noted by Demsetz and Villalonga (2001), an increase in Tobin Q may arise either because the firm improves operating performance or increases the present value of future investment opportunities. Monitoring by the board could either help in improving current performance or increase the value of investment opportunities, given the board's advisory role. We distinguish between these explanations by running additional tests using Return on Assets (ROA), a metric of the firm's operating performance, as the endogenous variable in our structural model.3 ROA is defined as the ratio of Earnings before Interest, Depreciation and Amortization to the book-value of total assets, and we adjust for industry effects at the 2-digit SIC level. We find no relationship between increased monitoring and ROA, suggesting that the board's main function is to provide strategic advice in order to increase the value of investment opportunities.

3

We thank the anonymous referee for this suggestion.

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

535

The level of monitoring activity and the probability of making the Audit, Compensation, and Nominating committees 100% independent increase with the independence of the entire board and the decrease with the size of the committee. Prior performance does not seem to impact the probability that firms move to fully independent boards. We also find that the decision to adopt fully independent Audit, Compensation and Nominating Committees does not result in subsequent incremental performance improvements. In no case, however, do we find that board monitoring reduces firm value. There are two possible explanations for our latter set of results. One, because of the intense focus of the media, Congress and regulators, firms moved quickly to have independent Boards and committees. Thus, the level of independence of the Board committees may not capture the variation of monitoring across firms as compared to the level of activity as proxied by the number of meetings. A second potential interpretation is that a statistically non-significant relation between committee organizational changes and firm value could imply that firms are in equilibrium, and are maximizing firm value (see Palia, 2001). The remainder of this paper is organized as follows. In the next section, we review the literature and summarize the hypotheses to be tested. Section three discusses the data and methodology we use. Section four summarizes the results and Section five concludes. 2. Literature Review and Hypotheses In this section, we review the literature that addresses the impact of board characteristics, firm characteristics and firm events upon board activity. We also discuss the literature on the impact of board monitoring on firm value, especially in the context of external regulation, and develop the hypotheses that motivate our tests and empirical analysis. The level of board meetings has been used as a metric of board activity in the literature. Vafeas (1999) was perhaps the earliest work to examine the determinants of the frequency of annual board meetings. Vafeas also argues that the frequency of board meetings is an important board attribute that can have important implications for firm value. He examines the number of board meetings in a sample of 307 firms over the period from 1990 to 1994. Adams (2005) expands the use of board meetings by incorporating the number of meetings held by the various board committees for 352 companies in 1998. Adams (2005) uses the number of board and committee meetings as a metric for a greater degree of board monitoring. We extend these studies by examining meeting frequency based monitoring proxies over a broad sample of firms over the period from 1999 to 2005. Vafeas (1999) and Adams (2005) argue that firm performance is an important determinant of board activity. Poor prior performance may require a greater need for monitoring to turn around the firm and boards may face increased pressure to be seen as being engaged when the firm is performing poorly. Both studies find empirical support for an inverse relationship between board meetings and prior performance. Literature has posited that the board fulfills an advising role (see Adams, 2005) and provides strategic advice that increases the value of investment opportunities available to the firm. Vafeas (1999) argues that the frequency of board meetings may increase with firm complexity and growth opportunities. We would also expect boards to meet more often in response to corporate events such as mergers and acquisitions. We therefore extend the literature by incorporating the role of corporate events in determining the level of board monitoring. The independence of the directors on the board can be an important determinant of board activity and indeed has been a focus of much of the shareholder and regulatory activity, e.g. an important provision of the 2002 Sarbanes-Oxley Act has been to increase the independence of the Board of Directors.4 As shown by Raheja (2005), insiders on the board serve to facilitate the flow of information to the board and the costs of information acquisition increase when the boards have a larger percentage of independent directors. The need to access information by other channels and the increased efforts needed for information coordination implies that board activity would increase as a board becomes more independent, as posited by Vafeas (1999) and Raheja (2005). We also incorporate the role of other governance measures in determining the level of board monitoring. Board activity can be impacted when the CEO is also the Chair of the Board of Directors. Literature has seen the dual role of the CEO as a sign of entrenchment of the CEO (see Linck et al., 2008). Board monitoring is compromised when the CEO is entrenched leading to a lower level of board activity. On the other hand board activity can increase to counter the entrenchment. Further the information effect of having an insider on the board is also relevant as the Board can have easier access to information when the CEO serves as the chair of the board. Board activity will decrease, as there will be a better flow of information to the board when the CEO is Board Chair, requiring a lower level of board activity devoted to information acquisition. Studies have shown that independence of the board committees also influences their monitoring role and firms are under increased pressure to ensure that board committees are more independent. Klein (2002) finds that earnings management decreases with the independence of the Audit Committee. Karamanou and Vafeas (2005) find a positive association between the reliability of management earnings and Audit Committee independence. We extend these studies by incorporating the frequency of committee meetings, especially the number of meetings of the Audit Committee, and a move to make committees fully independent as a proxy for the level of board activity.5 Literature has theorized both a positive and a negative impact of external pressure and regulation on firm value. On the one hand studies that have found a positive impact of good governance as measured by board attributes suggest that requiring firms to 4 The literature has found that board independence has important consequences for the firm. Weisbach (1988) find that independent boards are more likely than other boards to replace poorly performing management. Byrd and Hickman (1992), and Cotter et al. (1997) demonstrate that independent boards are more likely to obtain greater merger bids for the target shareholders than non-independent boards. 5 Reeb and Upadhyay (2010) find that the delegation of authority by the board to its committees and the number of committees established are related to board size and board independence.

536

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

improve the monitoring by the board of directors can enhance shareholder value. Chhaochharia and Grinstein (2007) find that those firms that were not in compliance with respect to the board and board committee related independence requirements of the Sarbanes-Oxley regulations earn positive abnormal returns after the public announcement of the passage of that bill by Congress. On the other hand, externally imposed regulations can have unintended and negative consequences (see Hermalin and Weisbach, 2006). Much of the board energy and activity may be spent on ensuring firm's compliance with regulations as opposed to managing the firm. Meetings can also be used inefficiently by independent board members who interpret their role to play the devil's advocate on behalf of shareholders. We note that this effect is different from the notion of inefficiencies arising from board being co-opted by the CEO (see Mace, 1986 and Jensen, 1993). Further, if the impetus behind increased board activity were simply the need to comply with regulation and the fear of stockholder litigation, we would expect increases in board activity to have a negative impact on firm value, as the increased activity would detract management from focusing on running the firm. As in the literature we use Tobin's Q as a measure of firm value (see Yermack, 1996, and Palia, 2001). As Demsetz and Villalonga (2001) point out however, changes in Tobin Q could either represent changes in operating performance or changes in the value of investment opportunities available to the firm. We distinguish between these alternate interpretations of Tobin Q, by using an accounting metric of firm performance Return on Assets (ROA), as an additional endogenous variable in our structural model and compare the results. The analysis of board monitoring cannot be independent of mechanisms, such as incentive compensation contracts, to align the CEO and shareholders. In our empirical analysis we therefore incorporate the pay-performance sensitivity of the CEO. The evidence on the impact of PPS on board monitoring and firm value are mixed.6 We also recognize that board monitoring and CEO payperformance-sensitivity can be endogenous and we use structural model regressions with instrumental variables in our empirical analysis.7 Our choice of instrumental variables is motivated by theoretical considerations in the literature and we also run statistical tests for the validity and strength of the instruments that we use. 3. Data and Methodology In this section, we describe the data and the methodology we use to examine the determinants of board monitoring and its impact upon firm value. We build a sample of firm-year observations for which we have complete data from various sources. We obtain data on board meetings from EXECUCOMP and data on board size and composition from RiskMetrics Group and The Corporate Library for the seven years from 1999 to 2005. We obtain compensation data from EXECUCOMP, firm accounting data from COMPUSTAT, and stock returns data from CRSP. In reporting the data and summary statistics we use the sample of firms for which we have board, compensation and firm characteristic data for all our observations. We have a complete set of data for a broad sample of 5,228 firm-year observations, which is comparable to other data sets used in the literature. Table 1 reports mean (median) sample values and distributional characteristics of our variables for these 5,228 observations. Appendix A delineates our variables and their expected impact on the level of board monitoring. 3.1. Monitoring based on the Board and Audit committee annual meetings We estimate a three-equation structural fixed-effects panel model with equations for the level of Board Monitoring Activity, Firm Value and PPS, with firm fixed effects. Firm fixed effects control for changes in our board monitoring activity proxies arising from changes to the firm's observable and unobservable characteristics. The specifications we estimate differentiate between a vector of common control variables X and a vector of specific instruments Z for each endogenous variable. Specifically, Monitoring Activity = αMA + βMA Firm Value + ωMA PPS + ΩMA X + λMA ZMA + εMA

ð1Þ

Firm Value = αQ + βQ Monitoring Activity + ωQ PPS + ΩQ X + λQ ZQ + εQ

ð2Þ

PPS = αPPS + βPPS Firm Value + ωPPS Monitoring Activity + ΩPPS X + λPPS ZPPS + εPPS

ð3Þ

3.1.1. Endogenous Variables We use two proxies for the level of Board Monitoring Activity for the board as a whole. One, we use the logarithm of the annual number of board meetings (LogMEETINGS). In taking the log, we assume that the amount of monitoring activity is non-linear in that the efficacy of monitoring exhibits decreasing returns to scale. Two, we use the log of the product of the number of

6 Studies have found both a positive relationship between compensation sensitivity and outsiders on the board (see e.g. Lambert et al., 1991) and a negative relationship (see e.g. Weisbach, 1988; Kole, 1995, Boone et al., 2007). McConnell and Servaes (1990) find a non-monotonic relationship between managerial compensation and firm value. 7 Literature has extensively studied the endogeneity of corporate governance and firm value. Wintoki et al. (2010) discuss three sources of endogeneity. One, endogeneity can arise from omitted variables that impact on both governance and performance. Two, endogeneity can arise from the simultaneous determination between governance and firm value. Three, endogeneity can arise from a dynamic relationship between firm governance and firm value. As discussed by Palia (2001), a 2SLS structural model with fixed effects takes into account endogeneity arising from omitted variables and a simultaneity. We also note that we use lagged TOBIN Q as an instrument to control for the impact of prior performance on current performance which is in the spirit of controlling for dynamic endogeneity.

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

537

Table 1 Table 1 reports the summary statistics of our variables. MEETINGS is the number of annual board of director meetings. MONITOR is the product of the number of independent directors and annual board meetings. PPS is the dollar value change in the portfolio of stocks and options held by the CEO for a one percent change in equity value, stated in millions of dollars. Tobin Q is the ratio of the total market value of the firm to the book value of the firm's assets. ROA is the return on assets defined as EBITDA/Total Assets. ACQDUM is a dummy variable equal to one if the firm undergone an acquisition during the current fiscal year and zero, otherwise. TARGETDUM is a dummy variable that is equal to one if the company was acquired during the fiscal year, and zero, otherwise. RESTATEDUM is a dummy variable that is equal to one if the company announced a restatement of earnings during the current fiscal year, and zero, otherwise. CEO_CHAIR is a dummy variable identifying firms in which the CEO is also Chairman. BSIZE is the number of directors on the board. INDEP is the percentage of the directors that are independent. ASSETS is the level of assets in millions of dollars for the previous fiscal year. VOLATILITY is the annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year. LEVERAGE is the ratio of debt to total assets. R&D is the ratio of R&D to total assets. G-INDEX is a proxy for the level of shareholder rights as measured by the GompersIshii-Metric Governance Index. CEOAGE is the age of the CEO. TENURE is the tenure of CEO measured in years. CEOINNOM is a dummy variable identifying if the CEO is a member of the Nominating Committee. RETURNM is the holding period return less one for the prior fiscal year. RETURN2M is the holding period return less one for the year two years prior to the current fiscal year. DIRMTGFEE is the fee in thousands of dollars per board meeting given to the director. DUMDIRPPS is a dummy variable that is equal to one if the compensation for the independent members of the board of directors contains an equity component. There are 5,228 observations.

MEETINGS MONITOR PPS TOBIN Q ROA ACQDUM TARGETDUM RESTATEDUM CEO_CHAIR BSIZE INDEP ASSETS VOLATILITY LEVERAGE R&D G-INDEX CEOAGE TENURE CEOINNOM RETURNM RETURN2M DIRMTGFEE DUMDIRPPS

Mean

Median

7.2645 47.3219 1.1423 1.9518 0.1401 0.0691 0.0011 0.0480 0.6530 9.2967 0.6786 6352.9860 0.4462 0.2155 0.0308 9.4227 55.6507 7.6363 0.0824 0.1520 0.1813 1.1672 0.8673

7 42 0.2651 1.4849 0.1341 0 0 0 1 9 0.7 1620.949 0.399 0.2099 0 9 56 5.1110 0 0.0796 0.0844 1 1

Std 3.0582 29.4965 6.7177 1.4997 0.104 0.2536 0.0339 0.2138 0.4761 2.3966 0.1624 17629.6100 0.1915 0.1678 0.0609 2.5995 7.1383 7.4595 0.2751 0.5567 0.6266 0.9588 0.3393

Minimum

Maximum

1 3 0 0.1650 -1.1235 0 0 0 0 3 0.1 31.169 0.124 0 0 2 33 0 0 -0.9129 -0.9401 0 0

49 396 191.7015 19.7790 0.9175 1 1 1 1 21 1 410063 1.527 1.5204 1.1249 17 90 54.7836 1 7.3586 9.6639 7.5 1

independent directors and the number of times the board meets in the year (LogMONITOR).8 As shown in Table 1, the mean (median) number of annual board meetings is 7.26 (7). No firm had less than one meeting per year and the maximum number of times the board is 49. The mean (median) of time spent on monitoring, which is equal to the number of independent directors multiplied by the number of board meetings, is 47.32 (42). We use Industry-adjusted Tobin Q (TOBIN Q) to proxy for Firm Value. We calculate the firm's Tobin Q as the ratio of the total market value of the firm, defined as the market value of the equity plus the book value of total debt to the book value of the firm's total assets. To find our measure, we take the difference between the computed Tobin Q for the firm and the equal weighted average of the Tobin Q for firms in the same industry at the two-digit SIC level. The unadjusted mean (median) Tobin Q in our sample is 1.952 (1.485). We use the pay-for-performance sensitivity (PPS) of the CEO's compensation to proxy for the use of incentives to align the manager with shareholders. We measure PPS as the sum of the dollar value changes in the CEO's stock and options compensations for a one percent change in the aggregate value of the firm equity. The procedure we use is analogous to that employed by Core and Guay (2002) and adds up the PPS measures of new option grants, options granted prior to the current fiscal year and shares of stock held by the CEO. The mean (median) PPS for our sample is $1.1423 million ($0.2651 million). 3.1.2. Instrumental Variables Our instrumental variables are based upon theoretical and empirical arguments from prior literature and we statistically test their validity and strength. We use four different instrumental variables for our monitoring activity variables (Eq. 1). We posit that director incentives should have a positive impact upon the level of board activity. We therefore use DUMDIRPPS, which equals one if the compensation of the director contains an equity component and DIRMTGFEE, which is the director meeting fee. 86.73% of firms provide option grants for their directors. The average (median) meeting fee is $1,167 ($1,000). We also use two proxies for prior

8

It has been shown in the literature that the benefit of board monitoring is a function of the number of independent members (see e.g. Weisbach, 1988).

538

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

performance as instrumental variables for board monitoring activity motivated by Vafeas (1999) and Adams (2005). Specifically, we use the industry-adjusted annual common stock return (defined as the difference between the firm's stock returns and the corresponding 2-digit SIC industry average return) for the prior fiscal year, RETURNM, and the industry adjusted annual common stock return in the year which is two years prior to the current fiscal year, RETURN2M, as in Vafeas (1999). The mean (median) of unadjusted RETURNM is 15.2% (7.96%). The mean (median) of unadjusted RETURN2M is 18.13% (8.44%). For Eq. (2), where Q is used as a proxy for firm value, we use lagged Q as our instrumental variable as there is a firm-specific persistence in Q. Palia (2001) and Coles et al. (2006) have also used lagged Q as an instrumental variable. We use the length of time that the CEO has been on the job, denoted as TENURE, as an instrumental variable for PPS (Eq. 3) as in Palia (2001). TENURE proxies for the bargaining power of the CEO with respect to the Board, as we expect longer serving CEOs to have greater bargaining power. The average (median) number of years the CEO is in office is 7.64 (5.11). We also use the age of the CEO, denoted as CEOAGE, as an instrumental variable for PPS. The average (median) age of the CEO is 55.65 (56). These variables capture the interaction between the manager's career concerns and incentive compensation. As Gibbons and Murphy (1992) argue there is a greater need to align managerial interest with that of shareholders the closer the manager is to retirement. We check for the validity and strength of our instruments using a set of econometric tests.9 We test whether our specifications are over-identified using a Hansen-Sargan test for over-identifying restrictions. We use the Cragg-Donald's minimum eigenvalue statistic to test for instrument relevance in model identification. We examine Shea's partial R2 and the associated F-test of the exclusion of the instrument set in the first-stage regression. We use the Anderson-Rubin statistic to examine the joint significance of the multiple endogenous variables in our system of equations. 3.1.3. Control Variables Our choice of control variables is motivated by their potential relevance as noted in prior literature (see Palia, 2001 and Vafeas, 1999). We classify our control variables into three different categories: Corporate Events, Governance Measures, and Scale and Complexity of the Firm. We expect an increase in the need for board monitoring in the fiscal years in which firms engage in merger activity or the restatement of earnings.10 We therefore include ACQDUM, TARGETDUM, and RESTATEDUM to account for these events. ACQDUM, is set equal to one if the firm acquired another firm during that fiscal year and is otherwise zero. Approximately, 6.91% of our sample firms were bidders in a given fiscal year. TARGETDUM, which is a dummy variable equal to one if the firm announced that they were acquired during the fiscal year, and zero otherwise. Only 0.11% of our sample firms were targets of an acquisition in a given fiscal year. RESTATEDUM, is equal to one if the firm restated its previously published earnings during the current fiscal year, and zero otherwise. Approximately 4.80% of our sample restated their earnings in a given fiscal year. We expect monitoring to increase with the scale of the firm. We proxy for the scale of the firm by the logarithm of the firm's total assets at the end of the firm's prior fiscal year, denoted as LogASSETS. The need for monitoring is also perhaps greater when the firm's operations are complex. On the other hand, however, the information requirements are greater the more complex the firm and monitoring can be lower. We use three proxies for firm complexity - VOLATILITY, Leverage and R&D. VOLATILITY is the volatility of stock returns and is measured as the annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year, as reported by EXECUCOMP. We use the firm's long-term debt divided by total assets as a measure of firm's LEVERAGE. R&D is the level of research and development expenses scaled by the total assets of the firm. We include several measures that represent the firm's corporate governance structure as control variables. Board size (BSIZE) and the percentage of independent directors (INDEP), impact upon the information coordination problem. These variables have also been shown to impact upon firm value (Yermack, 1996) and PPS (Coles et al., 2006). The mean (median) of BSIZE is 9.3 (9) and the mean (median) of INDEP is 67.86% (70%). We control for an entrenched CEO by including a dummy variable, denoted as CEO_CHAIR, which is equal to one when the CEO is also Chair of the Board of Directors. 65.3% of 5,228 firm observations have the CEO serving both roles. Additionally, we include a measure for shareholder rights, G-Index, as defined by the Gompers et al. (2003). The mean (median) G-Index in our sample is 9.43 (9.0). Finally, we included CEOINNOM, which equals one if the CEO is a member of the Nominating Committee or if the firm does not have a Nominating Committee. We assume that the level of management entrenchment is positively related to the ability of the CEO to pick the members of the board. Approximately 8.24% of our firms have the CEO as a member of the firm's Nominating Committee. 3.1.4. The Impact of Board Monitoring A positive coefficient for our board monitoring activity proxies in Eq. (2) indicates that board monitoring has a salutary impact. There are two possible ways in which the board impact upon firm value. One reason is that the board of directors helps in improving current performance. A second reason is that the board provides strategic advice that increases the value of investment opportunities presented to the firm. In order to discern between these two possibilities, we include an alternate performance measure, industry adjusted Return on Assets (ROA). The firm's unadjusted ROA is defined as the ratio of Earnings before Interest, Depreciation and Amortization to the book-value of total assets. The mean (median) non-industry adjusted ROA is 14.01% (13.41%). If the main impact of increased monitoring is to increase current performance then we should see similar positive

9

For a fuller discussion of the tests for instrumental variables, see Davidson and McKinnon (2004) and Brick et al. (2008). We expect that firm value and CEO compensation would be a function of particular corporate events such as a merger and acquisition or the realization of a required restatement of earnings (see, for example, Burns and Kedia, 2006 and Dennis et al., 2006). 10

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

539

relationship between Tobin Q and monitoring and ROA and monitoring. However, it is possible that the main impact of the board is to help identify valuable investment opportunities. In this case, there should still be a positive relationship between increased monitoring and Tobin Q but no relationship between increased monitoring and ROA.11 3.1.5. The Impact of Sarbanes-Oxley Our data sample period straddles the passage of the 2002 Sarbanes-Oxley Act. The act substantially increased the accountability of the board of directors and imposed independence requirements on board committees. Although the Act did not impose requirements regarding the number of board meetings, it is interesting to examine whether this Act change the relationship between board meetings and firm value. We study this issue in two ways. First, we estimated our structural equations for two distinct sub-periods, the years prior to the Act and the years after its passage. The pre-SOX period is a three-year period from 1999-2001 and the post-SOX period is a threeyear period from 2003-2005. We do not include 2002 in our analysis, as that was the transition when the Sarbanes-Oxley Act was enacted. We also require for each of the sub-samples, firms have observations in all three years. We impose this requirement because we are using a fixed-effects model. As a result, the sample size for the pre-SOX period is 1,293 and the sample size for the post-SOX period is 1,758. Second, we estimate the impact of the Act using the entire sample by including a new unitary variable, SOX, which is equal to zero for the pre-SOX period and is equal to one for the post-SOX period. Because we do not include firms in year 2002 and we run fixed-effects using an unbalanced panel, our sample size for this regression is 4,345. 3.1.6. Board Committee Meetings A major element of the Sarbanes-Oxley Act is the increased emphasis on Audit Committees. We examine the monitoring role of the Audit Committee using proxies similar to our board proxies, namely, the logarithm of the number of meetings held by the Audit Committees (LogAUDMTGS) and the logarithm of the product of Audit Committee Meetings and the number of independent members of that committee (LogAUDMONITOR). The impact of the Audit Committee monitoring activities could be different from that of the full board because the Act formally requires this committee to be fully independent and that it must meet, at the minimum, of four times per year. The mean (median) number for Audit Committee Meetings in the pre-SOX period is 3 (4) and in the post-SOX period is 7.66 (7). The mean (median) number for Audit Committee Director days (Audit Committee Meetings * Number of Independent Directors on the Audit Committee) in the pre-SOX period is 12.96 (12) and in the post-SOX period is 27.32 (24) respectively. Both the proxies for the level of monitoring by the audit committee show a dramatic increase in the period following the passage of the Sarbanes-Oxley period. 3.2. Board Committee Changes Our committee monitoring variables are AUD100, COMP100 and NOM100 for the Audit, Compensation, and Nominating committee respectively. In each case, the variable is equal to one in the year in which the firm appoints all independent directors to the Committee and is zero if the Committee remains at less than 100% independent in the current year and in the subsequent year. We posit that a move to a fully independent committee or starting a new committee is associated with an increase the level of board monitoring. Interestingly, we find that a significant minority of our sample had less than 100% independent Audit Committee even after the passage of the Sarbanes Oxley Act in 2002. This likely reflects a more strict definition of independence by our data source (RiskMetrics Group) than the definition promulgated by the exchanges. Our definitions of these board committee variables and the availability of a full panel of data for our regressions determines the sample size for our analysis. We find 933 firm-year observations where the Audit Committee is less than 100% independent and are candidates to become fully independent in the subsequent year. We set the dummy variable AUD100 to be equal to zero for these firms. We set AUD100 to one for the 328 firms-years in which firms do switch to a 100% independent Audit Committee. These firms are subsequently dropped after the Audit Committee becomes 100% independent. We find 750 cases where the Compensation Committee is less than 100% independent. Of these 750 firm-year observations, there are 495 firm-years in which the firm continues to have a less than 100% independent Compensation Committees (COMP100 = 0) and 255 firm-years in which the firm makes the switch to a fully independent Compensation Committee (COMP100 = 1), We find 1,100 cases where the Nominating Committee is less than 100% independent. Of these 1,100 firm-year observations, there were 751 firm-years in which the firm continues to have a less than 100% independent Nominating Committee (NOM100 = 0) and 349 firm-years in which the firm makes the switch to a fully independent Nominating Committee (NOM100 = 1). By construction our committee monitoring variables are zero/one variables. We use a two-step approach that takes the binary nature of the dependent variable. In the first step, we run PROBIT regressions that predict the likelihood of firms making these changes. More formally, PROBIT (First-Step): Change to Fully Independent Committee = αCM + ΩCM XCM + λCM ZCM + εCM

11

We thank the anonymous referee for this suggestion.

ð4Þ

540

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

Table 2 Table 2 summarizes the results of the multivariate 2SLS fixed effects regressions. Our dependent variables are LogMEETINGS, the logarithm of the number of annual board of director meetings, and LogMONITOR, the logarithm of the product of the number of independent directors and annual board meetings. The independent variables are as follows: PPS is the fitted value from the first stage of regression, representing the dollar value change in the portfolio of stocks and options held by the CEO for a one percent change in equity value. TOBIN Q is the fitted value from the first stage of regression, representing the ratio of the total market value of the firm to the book value of the firm's assets adjusted for industry mean at the two-digit SIC level. ACQDUM is a dummy variable equal to one if the firm undergone an acquisition during the current fiscal year and zero, otherwise. TARGETDUM is a dummy variable that is equal to one if the company was acquired during the fiscal year, and zero, otherwise. RESTATEDUM is a dummy variable that is equal to one if the company announced a restatement of earnings during the current fiscal year, and zero, otherwise. CEO_CHAIR is a dummy variable identifying firms in which the CEO is also Chairman. BSIZE is the number of directors on the board. INDEP is the percentage of the directors that are independent. LogASSETS is the logarithm of the assets in millions of dollars for the previous fiscal year. VOLATILITY is the annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year. LEVERAGE is the ratio of debt to total assets. R&D is the ratio of R&D to total assets. G-INDEX is a proxy for the level of shareholder rights as measured by the Gompers-Ishii-Metric Governance Index. CEOINNOM is a dummy variable identifying if the CEO is a member of the Nominating Committee. RETURNM is the returns for the prior fiscal year, adjusted for the industry mean at the two-digit SIC level. RETURN2M is the return for the year two years prior to the current fiscal year, adjusted for the industry mean at the two-digit SIC level. DIRMTGFEE is the fee per board meeting given to the director. DUMDIRPPS is a dummy variable that is equal to one if the compensation for the independent members of the board of directors contains an equity component. The Anderson-Rubin test examines the null hypothesis that PPS and Tobin Q are jointly equal to zero. The Hansen-Sargan test of instrumental validity examines the null hypothesis that instruments are not correlated with the structural error term. The table also provides the Cragg-Donald test of model under-identification and Shea's Partial R2 of instrument strength. There are 5,228 observations. Significance is indicated using bold font. LogMEETINGS

PPS TOBIN Q ACQDUM TARGETDUM RESTATEDUM CEO_CHAIR BSIZE INDEP LogASSETS VOLATILITY LEVERAGE R&D G-INDEX CEOINNOM RETURNM RETURN2M DIRMTGFEE DUMDIRPPS Anderson Rubin Test: (p-value) Hansen Sargan Test: (p-value) Cragg-Donald Test: (p-value) Shea's Partial R2: (p-value of first-stage F-statistics) PPS TOBIN Q

LogMONITOR

Coef.

P N |z|

Coef.

P N |z|

-0.0370 0.0549 0.1368 0.4417 0.0855 -0.0022 -0.0037 0.0472 0.0877 -0.1401 -0.0274 0.0821 -0.0026 0.0090 -0.0260 -0.0076 0.0170 0.0378

0.057 0.019 0.000 0.001 0.000 0.905 0.395 0.381 0.000 0.032 0.638 0.701 0.709 0.647 0.004 0.311 0.061 0.059 0.0556 0.4096 0.0000

-0.0490 0.0646 0.1507 0.3876 0.0789 -0.0307

0.026 0.015 0.000 0.011 0.001 0.135

1.8041 0.1370 -0.2249 -0.0515 0.2181 0.0057 0.0363 -0.0240 -0.0055 0.0233 0.0540

0.000 0.000 0.002 0.437 0.370 0.466 0.105 0.021 0.519 0.024 0.018 0.0131 0.1482 0.0000

0.0000 0.0000

0.0000 0.0000

XCM is the set of independent control variables and ZCM is our set of instrumental variables.12 We use the full set of control variables used in Eq. (1) that examines the degree of board monitoring.13 The instrumental variables for our PROBIT regressions are RETURNM, RETURN2M, and COMMSIZE. RETURNM and RETURN2M are the industry-adjusted stock returns in the prior fiscal year and the fiscal year two years prior, as defined earlier. We would expect that poor performance would increase the propensity to increase monitoring by the board committees.14 COMMSIZE is a measure of the size of the committee and is set equal to the number of directors on the relevant sub-committee, (i.e., the Audit, Compensation, and Nominating committees, in the respective regressions). We expect that the probability of the firm changing committee composition to be fully independent is negatively related to committee size. Our hypothesis is that smaller committees are more effective in instituting change, consistent with the empirical findings of Yermack (1996). A negative coefficient is also consistent with firms finding it easier to make smaller committees fully independent. We model the firm's decision to adopt a fully independent committee as taking place at the beginning of the fiscal year. Therefore the decision to change committee structure is driven by firm characteristics and performance in the previous fiscal year

12

When NOM100 is the dependent variable we drop CEOINOM. The exception is TARGETDUM. We do not use TARGETDUM as there are no cases where the firm announces a takeover in the prior fiscal year and survives to the end of the current fiscal year and changes its committee structure. 14 Coles et al. (2008) find a negative relationship between prior performance and board independence. 13

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

541

Table 3 Table 3 reports the results of our multivariate 2SLS fixed effects regressions. The dependent variable is PPS, the dollar value change in the portfolio of stocks and options held by the CEO for a one percent change in equity value. Our independent variables are as follows: LogMEETINGS is the fitted value from the first stage regression, representing the logarithm of the number of annual board of director meetings. LogMONITOR is the fitted value from the first stage regression, representing the logarithm of the product of the number of independent directors and annual board meetings. TOBIN Q is the fitted value from the first stage regression, representing the ratio of the total market value of the firm to the book value of the firm's assets adjusted for industry mean at the two-digit SIC level. ACQDUM is a dummy variable equal to one if the firm undergone an acquisition during the current fiscal year and zero, otherwise. TARGETDUM is a dummy variable that is equal to one if the company was acquired during the fiscal year, and zero, otherwise. RESTATEDUM is a dummy variable that is equal to one if the company announced a restatement of earnings during the current fiscal year, and zero, otherwise. CEO_CHAIR is a dummy variable identifying firms in which the CEO is also Chairman. BSIZE is the number of directors on the board. INDEP is the percentage of the directors that are independent. LogASSETS is the logarithm of the assets in millions of dollars for the previous fiscal year. VOLATILITY is the annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year. LEVERAGE is the ratio of debt to total assets. R&D is the ratio of R&D to total assets. G-INDEX is a proxy for the level of shareholder rights as measured by the Gompers-Ishii-Metric Governance Index. CEOINNOM is a dummy variable identifying if the CEO is a member of the Nominating Committee. CEOAGE is the age of the CEO. TENURE is the tenure of CEO measured in years. The Anderson-Rubin test examines the null hypothesis that LogMEETINGS (LogMONITOR) and Tobin Q are jointly equal to zero. The Hansen-Sargan test of instrumental validity examines the null hypothesis that instruments are not correlated with the structural error term. The table also provides the Cragg-Donald test of model under-identification and Shea's Partial R2 of instrument strength. There are 5,228 observations. Significance is indicated using bold font. Coef. LogMEETINGS LogMONITOR TOBIN Q ACQDUM TARGETDUM RESTATEDUM CEO_CHAIR BSIZE INDEP LogASSETS VOLATILITY LEVERAGE R&D G-INDEX CEOINNOM CEOAGE TENURE Anderson Rubin Test: (p-value) Hansen Sargan Test: (p-value) Cragg-Donald Test: (p-value) Shea's Partial R2: (p-value of first-stage F-statistics) LogMEETINGS LogMONITOR TOBIN Q

P N |z|

1.742

0.493

0.662 0.359 -0.961 0.005 0.378 -0.037 -1.151 -0.062 -0.600 -0.327 1.181 0.087 -0.013 -0.026 0.079

0.000 0.283 0.578 0.985 0.016 0.388 0.038 0.831 0.392 0.564 0.573 0.184 0.947 0.069 0.000 0.0013 0.6109 0.0018

Coef.

P N |z|

0.945 0.665 0.440 -0.527 0.081 0.410

0.658 0.000 0.155 0.733 0.745 0.031

-2.722 -0.055 -0.599 -0.321 1.132 0.077 -0.043 -0.026 0.077

0.499 0.876 0.428 0.569 0.593 0.242 0.833 0.086 0.000 0.0013 0.5412 0.0007

0.0036 0.0000

0.0013 0.0000

rather than in the current year. Consequently, the values for our firm characteristic and performance control variables are lagged one year relative to our binary Committee Monitoring proxy. However, the board and governance characteristics (e.g., CEO_CHAIR, BSIZE, INDEP, COMMSIZE, G-INDEX, and CEOINOM) are contemporaneous to that of the dependent variable because these are the board and governance characteristics at the beginning of the fiscal year when the board decides whether or not to change the committee structure. The second step in our analysis is the estimation of a two-equation structural model for the relationship between Tobin Q and PPS, while controlling for the firm's decision to adopt fully independent committees by including the Inverse Mills ratio computed from the PROBIT regressions in the structural 2SLS specification. More formally, we now estimate the following set of equations: Two-Stage Least Squares (Second-Step): Firm Value = αQ + βQ Inverse Mills Ratio + ωQ PPS + ΩQ X + λQ ZQ + εQ

ð5Þ

PPS = αPPS + βPPS Firm Value + ωPPS Inverse Mills Ratio + ΩPPS X + λPPS ZPPS + εPPS

ð6Þ

The values of these independent variables in these tests are contemporaneous to our three board dummy variables. As before, the instrumental variable for Tobin Q is LAGQ and the instrumental variables for PPS are CEOAGE and TENURE. The predictions with respect to the coefficient βQ are as follows. If the external pressure from stockholder activists and regulators shift the bargaining power away from entrenched managers to the board of directors, then βQ should be significantly positive. If the increased pressure simply increases the cost of monitoring, then βQ should be significantly negative. Finally, if firms are in equilibrium and are maximizing firm value, then we would expect βQ to be statistically insignificant (Palia, 2001).

542

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

Table 4 Table 4 reports the results of our multivariate 2SLS fixed effects regressions. The dependent variable is TOBIN Q, which is the ratio of the total market value of the firm to the book value of the firm's assets adjusted for industry mean at the two-digit SIC level. Our independent variables are as follows: LogMEETINGS is the fitted value from the first stage regression, representing the logarithm of the number of annual board of director meetings. LogMONITOR is the fitted value from the first stage regression, representing the logarithm of the product of the number of independent directors and annual board meetings. PPS is the fitted value from the first stage of regression, representing the dollar value change in the portfolio of stocks and options held by the CEO for a one percent change in equity value. ACQDUM is a dummy variable equal to one if the firm undergone an acquisition during the current fiscal year and zero, otherwise. TARGETDUM is a dummy variable that is equal to one if the company was acquired during the fiscal year, and zero, otherwise. RESTATEDUM is a dummy variable that is equal to one if the company announced a restatement of earnings during the current fiscal year, and zero, otherwise. CEO_CHAIR is a dummy variable identifying firms in which the CEO is also Chairman. BSIZE is the number of directors on the board. INDEP is the percentage of the directors that are independent. LogASSETS is the logarithm of the assets in millions of dollars for the previous fiscal year. VOLATILITY is the annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year. LEVERAGE is the ratio of debt to total assets. R&D is the ratio of R&D to total assets. G-INDEX is a proxy for the level of shareholder rights as measured by the Gompers-Ishii-Metric Governance Index. CEOINNOM is a dummy variable identifying if the CEO is a member of the Nominating Committee. LAGQ is the lagged value of TOBIN Q. The Anderson-Rubin test examines the null hypothesis that LogMEETINGS (LogMONITOR) and PPS are jointly equal to zero. The Hansen-Sargan test of instrumental validity examines the null hypothesis that instruments are not correlated with the structural error term. The table also provides the CraggDonald test of model under-identification and Shea's Partial R2 of instrument strength. There are 5,228 observations. Significance is indicated using bold font. Coef. LogMEETINGS LogMONITOR PPS ACQDUM TARGETDUM RESTATEDUM CEO_CHAIR BSIZE INDEP LogASSETS VOLATILITY LEVERAGE R&D G-INDEX CEOINNOM LAGQ Anderson Rubin Test: (p-value) Hansen Sargan Test: (p-value) Cragg-Donald Test: (p-value) Shea's Partial R2: (p-value of first-stage F-statistics) LogMEETINGS LogMONITOR PPS

P N |z|

1.478

0.060

0.008 -0.286 -0.853 -0.193 -0.009 -0.017 0.183 -0.618 -0.011 0.147 0.571 0.099 -0.126 0.268

0.899 0.015 0.133 0.041 0.882 0.266 0.334 0.000 0.964 0.458 0.441 0.000 0.057 0.000 0.0839 0.2569 0.0014

Coef.

P N |z|

1.186 0.016 -0.266 -0.640 -0.157 0.028

0.060 0.810 0.014 0.206 0.057 0.666

-1.870 -0.656 0.067 0.165 0.432 0.088 -0.161 0.265

0.113 0.000 0.802 0.397 0.560 0.000 0.016 0.000 0.0858 0.2309 0.0009

0.0008 0.0001

0.0001 0.0001

4. Empirical Results In this section we examine the results of our structural model regressions relating board monitoring by the board, firm value, and CEO PPS. Section 4.1 presents the results for the annual meeting monitoring proxies and Section 4.2 presents the results for changes in the composition of the key board committees. 4.1. Monitoring by the Full Board Tables 2, 3 and 4 present the 2SLS fixed effect regression results for monitoring by the full board for the three equations with LogMEETINGS/LogMONITOR, PPS, and TOBIN Q as the dependent variables. Each table presents 2 models. The first model uses LogMEETINGS as our proxy for monitoring while the second model uses LogMONITOR as our proxy for monitoring. As Table 2 shows, the determinants of board monitoring are similar for both our proxy variables, but are sharper for LogMONITOR. We find that board monitoring increases as the industry-adjusted Tobin Q increases. As the level of investment opportunities increases, the board's advising and monitoring activity increases. We find that PPS and Monitoring are negatively related and are, therefore, governance substitutes. With respect to our control variables, we find that the level of monitoring increases with corporate events as measured by ACQDUM, TARGETDUM and RESTATEDUM. We find that LogMONITOR increases with board monitoring as measured by INDEP, but LogMEETINGS is not related to INDEP. The positive relationship with LogMONITOR is consistent with increased coordination requirements when the board is more independent. There may also be a mechanical relationship, as LogMONITOR will be higher for boards of similar size but a higher proportion of independent directors. Monitoring activity also increase with changes in LogASSETS, suggesting that board activity is related to the cost of information coordination and the firm's scope. We also find that monitoring decreases with VOLATILITY, which is consistent with the predictions of Prendergast (2000) and Brick and Chidambaran (2008), who argue that monitoring activity may be less precise, and therefore less desirable, in uncertain

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

543

Table 5 Table 5 summarizes the results of the multivariate 2SLS fixed effects regressions. The endogenous variables are LogMEETINGS, which is the logarithm of the number of annual board of director meetings, PPS, which is the dollar value change in the portfolio of stocks and options held by the CEO for a one percent change in equity value, and ROA, which is the industry-adjusted return on assets defined as EBITDA/Total Assets. The independent variables are as follows. ACQDUM is a dummy variable equal to one if the firm undergone an acquisition during the current fiscal year and zero, otherwise. TARGETDUM is a dummy variable that is equal to one if the company was acquired during the fiscal year, and zero, otherwise. RESTATEDUM is a dummy variable that is equal to one if the company announced a restatement of earnings during the current fiscal year, and zero, otherwise. CEO_CHAIR is a dummy variable identifying firms in which the CEO is also Chairman. BSIZE is the number of directors on the board. INDEP is the percentage of the directors that are independent. LogASSETS is the logarithm of the assets in millions of dollars for the previous fiscal year. VOLATILITY is the annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year. LEVERAGE is the ratio of debt to total assets. R&D is the ratio of R&D to total assets. G-INDEX is a proxy for the level of shareholder rights as measured by the Gompers-Ishii-Metric Governance Index. CEOINNOM is a dummy variable identifying if the CEO is a member of the Nominating Committee. RETURNM is the returns for the prior fiscal year, adjusted for the industry mean at the two-digit SIC level. RETURN2M is the return for the year two years prior to the current fiscal year, adjusted for the industry mean at the two-digit SIC level. DIRMTGFEE is the fee per board meeting given to the director. DUMDIRPPS is a dummy variable that is equal to one if the compensation for the independent members of the board of directors contains an equity component. CEOAGE is the age of the CEO. TENURE is the tenure of CEO measured in years. LAGQ is the lagged value of TOBIN Q. The Anderson-Rubin test examines the null hypothesis that the endogenous variables are jointly equal to zero. The Hansen-Sargan test of instrumental validity examines the null hypothesis that instruments are not correlated with the structural error term. The table also provides the Cragg-Donald test of model under-identification and Shea's Partial R2 of instrument strength. There are 5,221 observations. Significance is indicated using bold font. LogMEETINGS Coef. PPS -0.045 ROA 2.632 LogMEETINGS ACQDUM 0.163 TARGETDUM 0.549 RESTATEDUM 0.090 CEO_CHAIR 0.012 BSIZE -0.006 INDEP 0.029 LogASSETS 0.119 VOLATILITY 0.029 LEVERAGE -0.150 R&D -0.404 G-INDEX 0.005 CEOINNOM 0.007 RETURNM -0.058 RETURN2M -0.015 DIRMTGFEE 0.027 DUMDIRPPS 0.040 CEOAGE TENURE LAGQ Anderson Rubin Test: (p-value) Hansen Sargan Test: (p-value) Cragg-Donald Test: (p-value) Shea's Partial R2: (p-value of first-stage F-statistics) PPS ROA LogMEETINGS

PPS P N |z|

0.065 0.037 0.000 0.001 0.000 0.612 0.214 0.657 0.000 0.784 0.102 0.226 0.550 0.751 0.009 0.110 0.017 0.074

0.0563 0.4934 0.0005 0.0000 0.0000

ROA

Coef.

P N |z|

28.606 3.055 0.461 -0.371 -0.068 0.528 -0.054 -1.426 0.202 1.483 -1.610 -4.217 0.166 -0.028 -0.375 -0.053

0.001 0.389 0.335 0.869 0.851 0.004 0.257 0.029 0.635 0.133 0.033 0.149 0.017 0.098 0.055 0.586

-0.041 0.077

0.851 0.000 0.0002 0.2180 0.0050

Coef.

P N |z|

–0.001

0.766

–0.065 -0.003 -0.018 0.002 -0.006 0.000 0.017 -0.020 -0.079 0.047 0.205 0.000 -0.002 0.010 0.001

0.252 0.712 0.590 0.695 0.078 0.942 0.083 0.000 0.000 0.000 0.000 0.741 0.609 0.000 0.564

0.007

0.000 0.3382 0.2449 0.0240 0.0000

0.0000 0.0077

0.0041

environments. Monitoring activity is inversely related to past firm performance as measured by RETURNM, consistent with the results reported by Vafeas (1999) and Adams (2005). As expected, the pressure on the board to meet and be seen to be active increases if the firm is doing poorly. The level of board monitoring activity also increases with the level of the meeting fee and whether the director's compensation includes an equity component. Finally, the statistical tests for instrumental variables reported in Table 2 imply that our system of equations is endogenous but not under-identified, and our instruments are well identified and strong. As Table 3 shows, we find no statistical relationship between PPS and board monitoring activity in any of the specifications. PPS increases with changes in Tobin Q. We also find that PPS is related to governance characteristics. PPS is positively associated with managerial entrenchment as proxied by TENURE and CEO_CHAIR. We also find that PPS declines as the CEO gets older. The statistical tests on instruments indicate that our system of equations is endogenous but not under-identified, and our instruments are well identified and strong. As shown in Table 4 there is a positive relationship between our board monitoring and Tobin Q implying that firm value increases with the level of board activity. These results are consistent with the notion that increased pressure from shareholder activists and regulator shift the bargaining power from management to shareholder, enhancing shareholder value. We do not find a significant positive relationship between PPS and Tobin Q, a result consistent with Palia (2001). In addition, merger and restatement of earnings activities decrease firm value. These results are consistent with the notion that low Tobin Q firms engage in merger activity or were forced to restate accounting earnings. Our regressions also imply that relative firm value decreases with

544

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

Table 6 Table 6 summarizes the results of the multivariate 2SLS fixed effects regressions. The endogenous variables are LogMONITOR, the logarithm of the product of the number of independent directors and annual board meetings, PPS, which is the dollar value change in the portfolio of stocks and options held by the CEO for a one percent change in equity value, and ROA, which is the industry-adjusted return on assets defined as EBITDA/Total Assets. The independent variables are as follows. ACQDUM is a dummy variable equal to one if the firm undergone an acquisition during the current fiscal year and zero, otherwise. TARGETDUM is a dummy variable that is equal to one if the company was acquired during the fiscal year, and zero, otherwise. RESTATEDUM is a dummy variable that is equal to one if the company announced a restatement of earnings during the current fiscal year, and zero, otherwise. CEO_CHAIR is a dummy variable identifying firms in which the CEO is also Chairman. BSIZE is the number of directors on the board. INDEP is the percentage of the directors that are independent. LogASSETS is the logarithm of the assets in millions of dollars for the previous fiscal year. VOLATILITY is the annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year. LEVERAGE is the ratio of debt to total assets. R&D is the ratio of R&D to total assets. G-INDEX is a proxy for the level of shareholder rights as measured by the Gompers-Ishii-Metric Governance Index. CEOINNOM is a dummy variable identifying if the CEO is a member of the Nominating Committee. RETURNM is the returns for the prior fiscal year, adjusted for the industry mean at the two-digit SIC level. RETURN2M is the return for the year two years prior to the current fiscal year, adjusted for the industry mean at the two-digit SIC level. DIRMTGFEE is the fee per board meeting given to the director. DUMDIRPPS is a dummy variable that is equal to one if the compensation for the independent members of the board of directors contains an equity component. CEOAGE is the age of the CEO. TENURE is the tenure of CEO measured in years. LAGQ is the lagged value of TOBIN Q. The Anderson-Rubin test examines the null hypothesis that the endogenous variables are jointly equal to zero. The Hansen-Sargan test of instrumental validity examines the null hypothesis that instruments are not correlated with the structural error term. The table also provides the Cragg-Donald test of model under-identification and Shea's Partial R2 of instrument strength. There are 5,221 observations. Significance is indicated using bold font. LogMONITOR Coef. PPS -0.059 ROA 3.126 LogMONITOR ACQDUM 0.182 TARGETDUM 0.517 RESTATEDUM 0.084 CEO_CHAIR -0.012 INDEP 1.785 LogASSETS 0.172 VOLATILITY -0.023 LEVERAGE -0.198 R&D -0.368 G-INDEX 0.014 CEOINNOM 0.033 RETURNM -0.062 RETURN2M -0.015 DIRMTGFEE 0.035 DUMDIRPPS 0.057 CEOAGE TENURE LAGQ Anderson Rubin Test: (p-value) Hansen Sargan Test: (p-value) Cragg-Donald Test: (p-value) Shea's Partial R2: (p-value of first-stage F-statistics) PPS ROA LogMEETINGS

PPS P N |z|

0.034 0.031 0.000 0.005 0.002 0.660 0.000 0.000 0.854 0.061 0.337 0.123 0.191 0.014 0.176 0.007 0.028

0.0133 0.2286 0.0005 0.0000 0.0000

ROA

Coef.

P N |z|

29.833 1.996 0.580 0.302 0.044 0.607 -4.846 0.193 1.663 -1.654 -4.694 0.146 -0.029 -0.423 -0.071

0.000 0.443 0.138 0.870 0.879 0.006 0.322 0.667 0.112 0.027 0.106 0.038 0.101 0.017 0.440

-0.098 0.076

0.669 0.000 0.0002 0.2153 0.0003

Coef.

P N |z|

–0.001

0.868

–0.039 -0.006 -0.031 0.000 -0.007 0.085 -0.020 -0.079 0.047 0.207 0.000 -0.001 0.010 0.001

0.308 0.368 0.245 0.979 0.045 0.234 0.000 0.000 0.000 0.000 0.926 0.837 0.000 0.373

0.007

0.000 0.3448 0.2015 0.0035 0.0000

0.0000 0.0007

0.0001

the level of assets. Interestingly, the coefficients on the G-Index are significantly positive, implying that firms that increase their GIndex have a higher Tobin Q. However, firm value is adversely affected if the CEO is on the Nominating Committee. Tobin Q is also positively associated prior performance as proxied by LAGQ. We note that Tobin Q is not related to board size, board independence, and CEO_CHAIR, consistent with the results of the dynamic endogeneity model of Wintoki et al. (2010). The statistical tests on instruments indicate that our system of equations is endogenous but not under-identified, that our instruments are well identified and strong.

4.2. ROA In this section, we present and discuss the empirical results when we use the industry-adjusted Return on Assets (ROA) as a measure of performance, instead of Tobin Q in the list of endogenous variables. Table 5 summarizes the results when LogMEETINGS is the monitoring proxy and Table 6 summarizes the results when LogMONITOR is the monitoring proxy.15 Unlike in Table 4 where changes in monitoring increase firm value, we find no relationship between board monitoring and ROA. This

15 Unlike Tables 2 – 4, RETURNM and RETURN2M are now treated as control variables. These variables cannot be instrumental variables since prior stock returns can anticipate the firm's current operating performance. Our statistical tests also indicate that these two variables cannot serve as instrumental variables.

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

545

Table 7 Table 7 reports the results of our multivariate 2SLS fixed effects regressions for split sample for years b 2002 and years N 2002. The dependent variable is TOBIN Q, which is the ratio of the total market value of the firm to the book value of the firm's assets adjusted for industry mean at the two-digit SIC level. Our independent variables are as follows: LogMEETINGS is the fitted value from the first stage regression, representing the logarithm of the number of annual board of director meetings. LogMONITOR is the fitted value from the first stage regression, representing the logarithm of the product of the number of independent directors and annual board meetings. PPS is the fitted value from the first stage of regression, representing the dollar value change in the portfolio of stocks and options held by the CEO for a one percent change in equity value. ACQDUM is a dummy variable equal to one if the firm undergone an acquisition during the current fiscal year and zero, otherwise. RESTATEDUM is a dummy variable that is equal to one if the company announced a restatement of earnings during the current fiscal year, and zero, otherwise. CEO_CHAIR is a dummy variable identifying firms in which the CEO is also Chairman. BSIZE is the number of directors on the board. INDEP is the percentage of the directors that are independent. LogASSETS is the logarithm of the assets in millions of dollars for the previous fiscal year. VOLATILITY is the annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year. LEVERAGE is the ratio of debt to total assets. R&D is the ratio of R&D to total assets. G-INDEX is a proxy for the level of shareholder rights as measured by the Gompers-Ishii-Metric Governance Index. CEOINNOM is a dummy variable identifying if the CEO is a member of the Nominating Committee. LAGQ is the lagged value of TOBIN Q. The Anderson-Rubin test examines the null hypothesis that LogMEETINGS (LogMONITOR) and PPS are jointly equal to zero. The Hansen-Sargan test of instrumental validity examines the null hypothesis that instruments are not correlated with the structural error term. The table also provides the Cragg-Donald test of model under-identification and Shea's Partial R2 of instrument strength. Significance is indicated using bold font. Years b 2002, N = 1,293 Coef. LogMEETINGS -8.386 LogMONITOR PPS 0.100 ACQDUM 0.501 RESTATEDUM -0.216 CEO_CHAIR 0.204 BSIZE 0.110 INDEP 0.231 LogASSETS -1.472 VOLATILITY 1.899 LEVERAGE 2.930 R&D 7.103 G-INDEX 0.223 CEOINNOM 0.144 LAGQ -0.221 Anderson Rubin Test: (p-value) Hansen Sargan Test: (p-value) Cragg-Donald Test: (p-value) Shea's Partial R2: (p-value of first-stage F-statistics) LogMEETINGS LogMONITOR PPS

P N |z|

Coef.

Years N 2002, N = 1,758 P N |z|

0.041 0.794 0.318 0.654 0.501 0.297 0.842 0.011 0.216 0.105 0.059 0.088 0.706 0.006 0.0000 0.0261 0.4313

-4.351 -0.120 0.078 -0.351 -0.075

0.123 0.645 0.819 0.303 0.690

9.075 -0.872 0.704 1.158 9.194 0.256 0.088 -0.166

0.081 0.007 0.531 0.271 0.002 0.027 0.745 0.010 0.0000 0.0000 0.5143

0.4632 0.0000

Coef.

P N |z|

0.516

0.305

-0.027 -0.175 -0.089 0.083 0.015 -0.062 -0.326 -0.248 0.222 -0.426 0.067 0.022 0.042

0.257 0.046 0.136 0.188 0.424 0.749 0.003 0.337 0.441 0.502 0.093 0.867 0.370 0.0030 0.0062 0.0108

Coef.

P N |z|

0.751 -0.033 -0.206 -0.094 0.106

0.149 0.198 0.027 0.136 0.133

-1.341 -0.379 -0.160 0.296 -0.332 0.062 0.055 0.052

0.151 0.003 0.575 0.328 0.627 0.144 0.699 0.296 0.0028 0.0206 0.0195

0.0144 0.6220 0.0000

0.0000

0.0127 0.0000

supports the notion that the main contribution of board monitoring is in helping identify investment opportunities as opposed to improving current performance.

4.3. The Impact of Sarbanes-Oxley Table 7 examines the impact of Sarbanes-Oxley upon firm value by re-estimating our structural model for the three-year period prior and for the three-year period after the passage of the Act.16 As Table 7 shows, the impact of monitoring upon firm value is negative in the first sub-period but it is positive and insignificant in the latter sub-period.17 We note, however, that the Hansen and Sargan statistics are significant implying that the instrumental variables are correlated with the structural error terms, and hence our coefficients may be biased. Table 8 summarizes our results for the full sample but including SOX as a control variable to capture the impact of the passage of the Sarbanes-Oxley Act in 2002. As the table shows, the coefficients for our monitoring proxies are positive, implying monitoring enhances firm value. In addition, the coefficient of SOX is also positively significant indicating that Sarbanes-Oxley had a positive impact upon firm value. 18

16

Note that we did not include 2002 in this analysis because the Act was discussed and promulgated during the calendar year. We also conducted our analysis with ROA as our dependent variable. We find no statistical relationship between board monitoring activity and operating performance. 18 We repeated the analysis with ROA as our dependent variable and find that monitoring does not affect ROA but Sarbanes-Oxley had a negative impact upon firm performance as measured by ROA. We interpret the negative impact upon ROA to be a reflection of the higher compliance costs arising from the 2002 Sarbanes-Oxley Act. 17

546

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

Table 8 Table 8 reports the results of our multivariate 2SLS fixed effects regressions. The dependent variable is TOBIN Q, which is the ratio of the total market value of the firm to the book value of the firm's assets adjusted for industry mean at the two-digit SIC level. Our independent variables are as follows: LogMEETINGS is the fitted value from the first stage regression, representing the logarithm of the number of annual board of director meetings. LogMONITOR is the fitted value from the first stage regression, representing the logarithm of the product of the number of independent directors and annual board meetings. PPS is the fitted value from the first stage of regression, representing the dollar value change in the portfolio of stocks and options held by the CEO for a one percent change in equity value. SOX is a dummy variable equal to zero for fiscal years 1999-2001 and one for fiscal years 2003–2005. ACQDUM is a dummy variable equal to one if the firm undergone an acquisition during the current fiscal year and zero, otherwise. TARGETDUM is a dummy variable that is equal to one if the company was acquired during the fiscal year, and zero, otherwise. RESTATEDUM is a dummy variable that is equal to one if the company announced a restatement of earnings during the current fiscal year, and zero, otherwise. CEO_CHAIR is a dummy variable identifying firms in which the CEO is also Chairman. BSIZE is the number of directors on the board. INDEP is the percentage of the directors that are independent. LogASSETS is the logarithm of the assets in millions of dollars for the previous fiscal year. VOLATILITY is the annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year. LEVERAGE is the ratio of debt to total assets. R&D is the ratio of R&D to total assets. G-INDEX is a proxy for the level of shareholder rights as measured by the Gompers-Ishii-Metric Governance Index. CEOINNOM is a dummy variable identifying if the CEO is a member of the Nominating Committee. LAGQ is the lagged value of TOBIN Q. The Anderson-Rubin test examines the null hypothesis that LogMEETINGS (LogMONITOR) and PPS are jointly equal to zero. The Hansen-Sargan test of instrumental validity examines the null hypothesis that instruments are not correlated with the structural error term. The table also provides the Cragg-Donald test of model under-identification and Shea's Partial R2 of instrument strength. There are 4,345 observations. Significance is indicated using bold font. Coef. LogMEETINGS LogMONITOR PPS SOX ACQDUM TARGETDUM RESTATEDUM CEO_CHAIR BSIZE INDEP LogASSETS VOLATILITY LEVERAGE R&D G-INDEX CEOINNOM LAGQ Anderson Rubin Test: (p-value) Hansen Sargan Test: (p-value) Cragg-Donald Test: (p-value) Shea's Partial R2: (p-value of first-stage F-statistics) LogMEETINGS LogMONITOR PPS

P N |z|

3.264

0.007

0.059 0.087 -0.568 -1.596 -0.395 0.025 -0.011 0.111 -0.873 0.030 0.043 0.866 0.081 -0.130 0.245

0.454 0.155 0.002 0.049 0.007 0.758 0.624 0.690 0.000 0.927 0.882 0.454 0.020 0.205 0.000 0.0005 0.3150 0.0081

Coef.

P N |z|

2.368 0.069 0.161 -0.491 -1.061 -0.302 0.092

0.008 0.351 0.002 0.001 0.107 0.010 0.279

-4.052 -0.956 0.158 0.190 0.624 0.050 -0.190 0.240

0.014 0.000 0.625 0.475 0.567 0.144 0.057 0.000 0.0005 0.1294 0.0022

0.0097 0.0000

0.0011 0.0000

4.4. Audit Committee Meetings In this section, we report the results describing the determinants and the impact of monitoring by the Auditing Committee. The two models reported in Table 9 summarize the results for LogAUDMTGS or LogAUDMONITOR respectively. Recall that the Sarbanes-Oxley Act of 2002 requires that the Audit Committee meet at least quarterly. This is clearly reflected by the positive coefficient for SOX. We find that as Tobin Q increases, the monitoring activity of the Audit Committee increases. The other results reported in Table 9 are generally analogous to those reported for the full board in Table 2. In particular, the monitoring activity of the Audit Committee is negatively related to prior performance and stock return volatility, and positively related to the percentage of independent directors and meeting fee of the directors. Unlike in Table 2, we find that Audit Committee activity increases as stockholder rights decrease. We note that the Hansen-Sargan test indicates that our instrumental variables are statistically valid for LogAUDMTGS but not for LogAUDMONITOR. Table 10 reports the impact of the Audit Committee's monitoring activity on firm value for LogAUDMTGS (Panel A) and LogAUDMONITOR (Panel B).19 We examine the impact of the monitoring activity in the full sample with SOX included as a control variable and for the pre-SOX and post-SOX sub-periods. We find that the monitoring activity of the AUDIT committee has a negative statistical impact upon firm value for the sample as whole. However, the results for the sub-periods provide an interesting contrast. In particular, there is a positive impact upon firm value for pre-SOX period but a (non-significant) negative impact in the post-Sox period. This suggests that the mandated minimum requirement for Audit Committee meetings impose unnecessary costs on the firm detracting from firm value. 19 The STATA fixed-effects subroutine, XTREG, dropped G-Index from the reported regression for the sub-sample prior to Sarbanes-Oxley because of collinearity problems.

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

547

Table 9 Table 9 summarizes the results of the multivariate 2SLS fixed effects regressions. Our dependent variables are LogAUDMTGS, the logarithm of the number of annual Audit Committee meetings, and LogAUDMONITOR, the logarithm of the product of the number of independent directors and annual Audit Committee meetings. The independent variables are as follows: PPS is the fitted value from the first stage of regression, representing the dollar value change in the portfolio of stocks and options held by the CEO for a one percent change in equity value. TOBIN Q is the fitted value from the first stage of regression, representing the ratio of the total market value of the firm to the book value of the firm's assets adjusted for industry mean at the two-digit SIC level. SOX is a dummy variable equal to zero for fiscal years 1999-2001 and one for fiscal years 2003-2005. ACQDUM is a dummy variable equal to one if the firm undergone an acquisition during the current fiscal year and zero, otherwise. TARGETDUM is a dummy variable that is equal to one if the company was acquired during the fiscal year, and zero, otherwise. RESTATEDUM is a dummy variable that is equal to one if the company announced a restatement of earnings during the current fiscal year, and zero, otherwise. CEO_CHAIR is a dummy variable identifying firms in which the CEO is also Chairman. BSIZE is the number of directors on the board. INDEP is the percentage of the directors that are independent. LogASSETS is the logarithm of the assets in millions of dollars for the previous fiscal year. VOLATILITY is the annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year. LEVERAGE is the ratio of debt to total assets. R&D is the ratio of R&D to total assets. G-INDEX is a proxy for the level of shareholder rights as measured by the Gompers-Ishii-Metric Governance Index. CEOINNOM is a dummy variable identifying if the CEO is a member of the Nominating Committee. RETURNM is the returns for the prior fiscal year, adjusted for the industry mean at the two-digit SIC level. RETURN2M is the annual return for the two years prior to the current fiscal year, adjusted for the industry mean at the two-digit SIC level. DIRMTGFEE is the fee per board meeting given to the director. DUMDIRPPS is a dummy variable that is equal to one if the compensation for the independent members of the board of directors contains an equity component. The Anderson-Rubin test examines the null hypothesis that PPS and Tobin Q are jointly equal to zero. The Hansen-Sargan test of instrumental validity examines the null hypothesis that instruments are not correlated with the structural error term. The table also provides the Cragg-Donald test of model under-identification and Shea's Partial R2 of instrument strength. There are 3,350 observations. Significance is indicated using bold font. LogAUDMTGS

PPS TOBIN Q SOX ACQDUM TARGETDUM RESTATEDUM CEO_CHAIR BSIZE INDEP LogASSETS VOLATILITY LEVERAGE R&D G-INDEX CEOINNOM RETURNM RETURN2M DIRMTGFEE DUMDIRPPS Anderson Rubin Test: (p-value) Hansen Sargan Test: (p-value) Cragg-Donald Test: (p-value) Shea's Partial R2: (p-value of first-stage F-statistics) PPS TOBIN Q

LogAUDMONITOR

Coef.

P N |z|

Coef.

P N |z|

-0.058 0.101 0.604 -0.006 0.050 0.017 0.015 -0.008 0.231 0.213 -0.278 -0.109 0.145 0.038 -0.056 -0.035 -0.066 0.051 -0.021

0.252 0.016 0.000 0.910 0.793 0.627 0.599 0.290 0.016 0.000 0.042 0.271 0.710 0.008 0.117 0.026 0.000 0.012 0.548 0.0259 0.1210 0.0002

-0.045 0.113 0.629 0.007 -0.030 -0.006 -0.037

0.442 0.022 0.000 0.907 0.894 0.881 0.283

1.563 0.284 -0.105 -0.122 0.248 0.040 -0.031 -0.041 -0.064 0.040 -0.005

0.000 0.000 0.512 0.301 0.589 0.017 0.463 0.027 0.000 0.092 0.907 0.0051 0.0139 0.0001

0.000 0.000

0.000 0.000

We also ran similar regressions using monitoring proxies based on the number of Compensation Committee meetings (LogCOMPMTGS and LogCOMPMONITOR) and the number of Nominating Committee meetings (LogNOMMTGS and LogNOMMONITOR). Our results (not reported) indicate that committee activity increased after the 2002 Sarbanes-Oxley Act but firm value is unrelated to these monitoring proxies. 4.5. Board Committee Changes We next examine the determinants in the changes in the composition structure of the Audit, Compensation and Nominating Committees and the impact of these changes on firm value. Models 1-3 of Table 11 report the results of the first step PROBIT regressions when AUD100, COMP100, and NOM100 are respectively the dependent variables. Tables 12 and 13 summarize the results of the second step structural model for firm value, Tobin Q and PPS. Model 1 in Table 11 shows the results of our PROBIT regression when the dependent variable is AUD100. As the table shows, the coefficient on the SOX dummy is significant and positive suggesting that the likelihood that a firm will implement a fully independent Audit Committee increased after the passage of the 2002 Sarbanes-Oxley Act. We find that the likelihood that AUD100 is 1 increases with the level of board independence (INDEP). There could be two explanations for this result. One, a more independent board can push to increase oversight by the Audit Committee by making it fully independent. Alternatively, a Board that has more independent directors may simply find it easier to construct committees that are fully independent. We also find

548

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

Table 10 Table 10 reports the results of our multivariate 2SLS fixed effects regressions for the full sample and split samples for years b 2002 and years N 2002. The dependent variable is TOBIN Q, which is the ratio of the total market value of the firm to the book value of the firm's assets adjusted for industry mean at the two-digit SIC level. Panel A reports results when we use the fitted value of LogAUDMTGS, representing the logarithm of the number of annual Audit Committee meetings, as the proxy for board monitoring activity. Panel B reports results when we use the fitted value of LogAUDMONITOR, representing the logarithm of the product of the number of independent directors on the Audit Committee and annual Audit Committee meetings, as the proxy for board monitoring activity. PPS is the fitted value from the first stage of regression, representing the dollar value change in the portfolio of stocks and options held by the CEO for a one percent change in equity value. SOX is a dummy variable equal to zero for fiscal years 1999-2001 and one for fiscal years 2003-2005. The control variables not reported but included in the regressions are as follows ACQDUM is a dummy variable equal to one if the firm undergone an acquisition during the current fiscal year and zero, otherwise. TARGETDUM is a dummy variable that is equal to one if the company was acquired during the fiscal year, and zero, otherwise. RESTATEDUM is a dummy variable that is equal to one if the company announced a restatement of earnings during the current fiscal year, and zero, otherwise. CEO_CHAIR is a dummy variable identifying firms in which the CEO is also Chairman. BSIZE is the number of directors on the board. INDEP is the percentage of the directors that are independent. LogASSETS is the logarithm of the assets in millions of dollars for the previous fiscal year. VOLATILITY is the annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year. LEVERAGE is the ratio of debt to total assets. R&D is the ratio of R&D to total assets. G-INDEX is a proxy for the level of shareholder rights as measured by the Gompers-Ishii-Metric Governance Index. CEOINNOM is a dummy variable identifying if the CEO is a member of the Nominating Committee. LAGQ is the lagged value of TOBIN Q. The Anderson-Rubin test examines the null hypothesis that LogAUDMTGS (LogAUDMONITOR) and PPS are jointly equal to zero. The Hansen-Sargan test of instrumental validity examines the null hypothesis that instruments are not correlated with the structural error term. The table also provides the Cragg-Donald test of model under-identification and Shea's Partial R2 of instrument strength. Significance is indicated using bold font. PANEL A

Full Sample N = 3,350 Coef.

LogAUDMTGS -0.613 PPS 0.013 SOX 0.376 Control Variables Yes Instruments Yes Anderson Rubin Test: (p-value) Hansen Sargan Test: (p-value) Cragg-Donald Test: (p-value) Shea's Partial R2: (p-value of first-stage F-statistics) LogAUDMTGS PPS PANEL B

LogAUDMONITOR -0.623 PPS 0.004 SOX 0.400 Control Variables Yes Instruments Yes Anderson Rubin Test: (p-value) Hansen Sargan Test: (p-value) Cragg-Donald Test: (p-value) Shea's Partial R2: (p-value of first-stage F-statistics) LogAUDMONITOR PPS

Years N 2002 N = 2,020

Coef.

P N |z|

Coef.

P N |z|

0.082 0.841 0.085

2.587 -0.263

0.079 0.523

-0.399 -0.109

0.416 0.271

Yes Yes

Yes Yes

0.4009 0.6213 0.0001

0.0559 0.7359 0.5521

0.0905 0.1524 0.0836

0.000 0.000

0.1920 0.3196

0.1270 0.0000

Full Sample N = 3,350 Coef.

Years b 2002 N = 1,019

P N |z|

Years b 2002 N = 1,019

Years N 2002 N = 2,020

P N |z|

Coef.

P N |z|

Coef.

P N |z|

0.073 0.950 0.075

2.208 0.000

0.133 0.999

-0.162 -0.110

0.607 0.258

Yes Yes

Yes Yes

0.3815 0.6875 0.0008

0.0470 0.4037 0.6265

0.0831 0.0978 0.0112

0.0012 0.0000

0.6909 0.3033

0.0160 0.0000

that the probability that AUD100 is equal to 1 is higher when the CEO is not the chair of the board, VOLATILITY is higher, and for non-R&D firms, The decision to have a fully independent Audit Committee is not statistically related to whether the firm restated their earnings or acquired a company in the previous year and, surprisingly, is also not related to prior stock performance. These results are contrary to our expectations, but imply that regulatory pressure, not firm performance, have been the prime drivers of changes in the Audit Committee. We find that the coefficient on committee size is negative and significant. Our finding lends empirical support for our hypothesis that boards find it easier to make smaller committees fully independent and that smaller committees are more effective in instituting change.20 These results could imply that firms find it easier to make the committees more independent when the board is more independent or when the committees are smaller. Alternatively, these results also imply that boards and committees that are less beholden to entrenched management have a higher likelihood of making committees more independent. We also note that the χ2 statistic indicates that the regression model is highly significant. 20 Committee size may increase or decrease simultaneously with changes in the composition of the committee, which can result in a mechanical relationship between changes in composition and committee size. We therefore re-run our test on the subset of firms that do not change committee size. In our sample, 550 of the 933 firms with AUD100 = 0 or AUD100 = 1, 479 of the 750 firms with COMP100 = 0 or COMP100 = 1, and 579 of the 1,100 firms with NOM100 = 0 or NOM100 = 1, have no changes in committee size. Our results, with respect to the sign and the statistical significance of the coefficients, remain the same as those reported in Tables 11-13 in these subsamples as well.

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

549

Table 11 Table 11 reports the results of PROBIT regressions to determine the factors that determine the probability that a firm switches to independent Committees. Model 1 presents the results for firms that did not have 100% independent Audit Committee by 1999. Model 2 presents the results for firms that did not have 100% independent Compensation Committee by 1999. Model 3 presents the results for firms that did not have 100% independent Nominating Committee by 1999. The dependent variables are AUDIT100, COMP100, and NOM100 respectively, which are set equal to one if the firm appoints all independent directors to the Committee in year t and it was not fully independent in year t-1, and is zero if the Committee remains at less than 100% independent at t. The independent variables are as follows: ACQDUM is a dummy variable equal to one if the firm undergone an acquisition during the previous fiscal year and zero, otherwise. RESTATEDUM is a dummy variable that is equal to one if the company announced a restatement of earnings during the prior fiscal year, and zero, otherwise. SOX is a dummy variable equal to zero for fiscal years 1999-2001 and one for fiscal years 2003-2005. CEO_CHAIR is a dummy variable identifying firms in which the CEO is also Chairman. BSIZE is the number of directors on the board. INDEP is the percentage of the directors that are independent. LogASSETS is the logarithm of the assets in millions of dollars for the previous fiscal year. VOLATILITY is the annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year. LEVERAGE is the ratio of debt to total assets. R&D is the ratio of R&D to total assets. G-INDEX is a proxy for the level of shareholder rights as measured by the Gompers-Ishii-Metric Governance Index. CEOINNOM is equal to one if the firm does not have a Nominating Committee or if the CEO is present on the Nominating Committee. RETURNM and RETURN2M are the industry-adjusted stock returns in fiscal years t-1 and t-2, respectively. COMMSIZE is the number of directors in the Audit, Compensation, and Nominating committees in Model 1, Model 2, and Model 3 respectively. The table also presents the log-likelihood and the χ2 statistics for the regression and reports the sample size for each model. Significance is indicated using bold font. Model 1

Model 2

Audit Committee

INTERCEPT ACQDUM RESTATEDUM SOX BSIZE INDEP CEO_CHAIR LogASSETS VOLATILITY LEVERAGE R&D G-INDEX CEOINNOM RETURNM RETURN2M COMMSIZE Log Likelihood χ2 Number of Obs.

Model 3

Compensation Committee

Nominating Committee

Coef.

P N |z|

Coef.

P N |z|

Coef.

P N |z|

-2.688 0.112 0.010 0.325 0.042 4.737 -0.175 -0.010 0.884 0.180 -2.754 -0.004 0.247 -0.010 0.104 -0.358 -483.24 243.45 933

0.000 0.548 0.968 0.001 0.125 0.000 0.084 0.800 0.005 0.572 0.014 0.853 0.166 0.909 0.202 0.000

-2.955 0.098 0.237 0.280 -0.050 5.021 -0.055 0.095 0.255 -0.007 -0.443 -0.020 0.086 0.019 0.103 -0.235 -368.95 223.65 750

0.000 0.691 0.383 0.021 0.106 0.000 0.628 0.054 0.454 0.984 0.716 0.362 0.686 0.838 0.279 0.000

-4.218 0.405 -0.026 1.051 0.020 5.638 -0.081 0.018 0.447 -0.517 -1.430 -0.038

0.000 0.050 0.915 0.000 0.463 0.000 0.421 0.640 0.132 0.120 0.174 0.053

-0.068 -0.047 -0.211 -499.52 375.51 1100

0.471 0.585 0.000

0.0000

0.000

0.000

Models 2 and 3 indicate that the regression results when the dependent variable is COMP100 and NOM 100 are largely similar to the results for the case when AUD100. The differences are as follows. Model 2 shows that the likelihood that the firm will adopt a fully independent Compensation Committee increases with firm size. Model 3 shows that the likelihood that a Nominating Committee will become 100% independent increases when the firm chooses to acquire another company and decreases with the G-Index. The regression has a high χ2 statistic indicating that it is highly significant. Our evidence is consistent with the argument that regulatory pressure has played an important role in changing committee structure. We next examine whether the move to a fully independent Audit, Compensation and Nominating Committees, affects Tobin Q and PPS. Models 1-3 of Table 12 present the results for the 2SLS model for the Audit, Compensation, and Nominating Committees respectively, when TOBIN Q is the dependent variable. The coefficients on LAMBDA in Table 12 are not significant for any of the models implying that the endogenous choice firms make to make any of the committees 100% independent has little impact upon firm value.21 There are three possible explanations for this. One, monitoring by the Board committees has no impact upon firm value. Second, because of the intense focus of the media, Congress and regulators, all firms moved quickly to increase the independence of Boards and its committees mitigating the cross-sectional impact on firm value. A third potential interpretation is that a statistically non-significant relation between committee organizational changes and firm value could imply that firms are in equilibrium, and are maximizing firm value (see Palia, 2001). We note that the coefficient for SOX is negative and significant. Our results indicate that those firms that chose not to move to 100% independent committees were the low performing firms as proxied by Tobin Q. Models 1-3 of Table 13 summarize our 2SLS results for the Audit, Compensation, and Nominating Committees respectively, when PPS is the dependent variable. We find that the coefficient for LAMBDA is negative and significant in Model 1 implying that firms choosing to make the audit committees 100% independent have a lower PPS. The negative relationship could arise because of 21 We also ran empirical tests where we use AUD100 as an independent variable in Eqs. (5) and (6), instead of the Inverse Mills Ratio. The coefficient was also not statistically significant. We similarly separately estimated Eqs. (5) and (6) using COMP100 and NOM100 with similar results.

550

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

Table 12 Table 12 presents the results for the first equation of a 2SLS model for Tobin Q and PPS for three different samples. Model 1 summarizes the sample for those firms that did not have 100% independent Audit Committee by 1999. Model 2 summarizes the sample for those firms that did not have 100% independent Compensation Committee by 1999. Model 3 summarizes the sample for those firms that did not have 100% independent Nominating Committee by 1999. The dependent variable is TOBIN Q, which is the ratio of the total market value of the firm to the book value of the firm's assets adjusted for industry mean at the two-digit SIC level. Our independent variables are as follows: PPS is the fitted value from the first stage of regression, representing the dollar value change in the portfolio of stocks and options held by the CEO for a one percent change in equity value. ACQDUM is a dummy variable equal to one if the firm undergone an acquisition during the prior fiscal year and zero, otherwise. RESTATEDUM is a dummy variable that is equal to one if the company announced a restatement of earnings during the prior fiscal year, and zero, otherwise. CEO_CHAIR is a dummy variable identifying firms in which the CEO is also Chairman. BSIZE is the number of directors on the board. INDEP is the percentage of the directors that are independent. LogASSETS is the logarithm of the assets in millions of dollars for the previous fiscal year. VOLATILITY is the annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year. LEVERAGE is the ratio of debt to total assets. R&D is the ratio of R&D to total assets. G-INDEX is a proxy for the level of shareholder rights as measured by the Gompers-Ishii-Metric Governance Index. CEOINNOM is a dummy variable identifying if the CEO is a member of the Nominating Committee. LAGQ is the lagged value of TOBIN Q. SOX is a dummy variable equal to zero for fiscal years 1999-2001 and one for fiscal years 2003-2005. LAMBDA is the Inverse Mills Ratio is from the first stage PROBIT regression for each of the committees respectively. The Anderson-Rubin test examines the null hypothesis that PPS is equal to zero. The Hansen-Sargan test of instrumental validity examines the null hypothesis that instruments are not correlated with the structural error term. The table also provides the Cragg-Donald test of model underidentification and Shea's Partial R2 of instrument strength. Significance is indicated using bold font. Model 1

Model 2

Audit Committee

PPS ACQDUM RESTATEDUM CEO_CHAIR BSIZE INDEP LogASSETS VOLATILITY LEVERAGE R&D G-INDEX CEOINNOM LAGQ SOX LAMBDA INTERCEPT Anderson Rubin Test Hansen Sargan Test Cragg-Donald Test Shea's Partial R2: for PPS Number of Obs.

Model 3

Compensation Committee

Nominating Committee

Coef.

P N |z|

Coef.

P N |z|

Coef.

P N |z|

0.070 -0.320 -0.059 0.093 -0.018 -0.030 -0.120 -0.952 0.208 3.186 0.014 0.059 0.584 -0.249 0.056 1.247

0.002 0.058 0.753 0.231 0.380 0.893 0.019 0.000 0.451 0.000 0.334 0.647 0.000 0.001 0.281 0.000 0.0004 0.4702 0.0000 0.0000 933

0.065 -0.055 -0.094 -0.045 -0.049 -0.018 -0.054 -0.984 0.286 2.630 0.021 0.082 0.559 -0.149 -0.041 1.007

0.005 0.781 0.567 0.491 0.004 0.928 0.108 0.000 0.210 0.000 0.106 0.500 0.000 0.032 0.395 0.000 0.0032 0.1422 0.0000 0.0000 750

0.017 -0.227 -0.031 -0.031 -0.026 0.159 0.000 -0.703 -0.211 3.336 -0.002 -0.042 0.616 -0.156 -0.033 0.544

0.225 0.031 0.784 0.478 0.018 0.284 0.993 0.000 0.173 0.000 0.834 0.520 0.000 0.001 0.292 0.005 0.0696 0.0482 0.0000 0.0000 1100

the relationship between PPS, Audit Committee Independence, and R&D activity. As shown in Table 11, low R&D firms are more likely to move to a fully independent Audit Committee, even after controlling for the effect of SOX. In addition Coles et al. (2008) show that low R&D firms have lower PPS. The negative coefficient on the Inverse Mills Ratio for the Audit Committee Regressions is therefore consistent with the findings of Coles et al. (2008). The coefficient on LAMBDA is positive and significant in Model 2 implying that firms choosing to make the compensation committees 100% independent have a higher PPS, implying that a more independent Compensation Committee will be more likely to use incentive compensation to motivate the CEO and align the CEO's interests with that of shareholders. We also find that PPS increases with Tobin Q, implying that managers in high value firms have greater pay-performance sensitivity. Third, we find that characteristics impact on PPS. PPS increases with firm size, but is negatively related to the leverage of the firm and if the CEO becomes Chairman of the Board. Finally PPS increases as firms become involve in acquisition activity and with increases with the tenure of the CEO. 5. Conclusions The recent regulatory and political pressure, starting with the 1998 Blue Ribbon Committee on Improving the Effectiveness of Corporate Audit Committees and the 2002 Sarbanes-Oxley Act, have served to make boards independent and accountable. Much of the attention on the Board of Directors and regulatory action has been predicated on the assumption that board monitoring activity can enhance shareholder value. In this paper, we examine the determinants of board activity, the impact of board activity on firm value and the impact of the Sarbanes-Oxley Act. Our sample, for which we have all the data for the endogenous variables, independent control variables and instrumental variables, consists of a broad panel of 5,228 firm-year observations over a six-year period from 1999 to 2005. We show that board activity is driven by prior performance as has been noted by Vafeas (1999) and Adams (2005). We also show that corporate events, such as a proposed merger or acquisition are strong determinants of board monitoring activity in our

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

551

Table 13 Table 13 presents the results for the second equation of the 2SLS model for Tobin Q and PPS for three different samples. Model 1 summarizes the sample for those firms that did not have 100% independent Audit Committee by 1999. Model 2 summarizes the sample for those firms that did not have 100% independent Compensation Committee by 1999. Model 3 summarizes the sample for those firms that did not have 100% independent Nominating Committee by 1999. The dependent variable is PPS, the dollar value change in the portfolio of stocks and options held by the CEO for a one percent change in equity value. Our independent variables are as follows: TOBIN Q is the fitted value from the first stage regression, representing the ratio of the total market value of the firm to the book value of the firm's assets adjusted for industry mean at the two-digit SIC level. ACQDUM is a dummy variable equal to one if the firm undergone an acquisition during the current fiscal year and zero, otherwise. RESTATEDUM is a dummy variable that is equal to one if the company announced a restatement of earnings during the current fiscal year, and zero, otherwise. CEO_CHAIR is a dummy variable identifying firms in which the CEO is also Chairman. BSIZE is the number of directors on the board. INDEP is the percentage of the directors that are independent. LogASSETS is the logarithm of the assets in millions of dollars for the previous fiscal year. VOLATILITY is the annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year. LEVERAGE is the ratio of debt to total assets. R&D is the ratio of R&D to total assets. G-INDEX is a proxy for the level of shareholder rights as measured by the Gompers-Ishii-Metric Governance Index. CEOINNOM is a dummy variable identifying if the CEO is a member of the Nominating Committee. CEOAGE is the age of the CEO. TENURE is the tenure of CEO measured in years. SOX is a dummy variable equal to zero for fiscal years 1999-2001 and one for fiscal years 2003-2005. LAMBDA is the Inverse Mills Ratio is from the first stage PROBIT regression for each of the committees respectively. The Anderson-Rubin test examines the null hypothesis that Tobin Q is equal to zero. The Hansen-Sargan test of instrumental validity examines the null hypothesis that instruments are not correlated with the structural error term. The table also provides the Cragg-Donald test of model under-identification and Shea's Partial R2 of instrument strength. Significance is indicated using bold font. Model 1

Model 2

Audit Committee

TOBIN Q ACQDUM RESTATEDUM CEO_CHAIR BSIZE INDEP LogASSETS VOLATILITY LEVERAGE R&D G-INDEX CEOINNOM CEOAGE TENURE SOX LAMBDA INTERCEPT Anderson Rubin Test Hansen Sargan Test Cragg-Donald Test Shea's Partial R2 for Q Number of Obs.

Model 3

Compensation Committee

Nominating Committee

Coef.

P N |z|

Coef.

P N |z|

Coef.

P N |z|

1.272 4.492 -0.633 -2.014 -0.412 0.341 1.964 2.021 -6.868 -1.873 -0.082 0.507 -0.078 0.231 0.653 -0.980 -4.663

0.000 0.000 0.707 0.004 0.011 0.864 0.000 0.329 0.001 0.748 0.511 0.655 0.108 0.000 0.305 0.021 0.192 0.0000 0.0000 0.0000 0.0000 933

0.807 5.180 -0.150 -0.835 0.069 -0.753 0.776 1.700 -3.175 0.085 -0.081 0.269 0.020 0.168 0.621 0.550 -6.503

0.001 0.000 0.899 0.111 0.583 0.595 0.000 0.267 0.035 0.984 0.377 0.758 0.574 0.000 0.208 0.095 0.016 0.0008 0.0000 0.0000 0.0000 750

0.764 3.267 0.351 -1.042 -0.101 -0.911 0.732 1.138 -3.228 -2.388 0.011 0.225 -0.013 0.231 0.217 0.173 -3.196

0.000 0.000 0.687 0.004 0.236 0.427 0.000 0.288 0.004 0.453 0.871 0.657 0.623 0.000 0.565 0.478 0.110 0.0001 0.0000 0.0000 0.0000 1100

sample. Our evidence indicates that regulatory pressure, specifically the passage of the Sarbanes-Oxley Act in 2002, has increased board activity. We also find that firm value as proxied by Tobin Q is higher when board monitoring is higher. This supports the notion that monitoring by the entire board leads to increase firm value consistent with the recent evidence in Chhaochharia and Grinstein (2007) and Vafeas (1999). We also find that board monitoring does not impact ROA, suggesting that the main contribution of board monitoring is in helping identify investment opportunities as opposed to improving current operating performance. Boards have also moved to making their Audit. Compensation, and Nominating committees fully independent, especially after the passage of the 2002 Sarbanes-Oxley Act, and the changes are more likely to occur with more independent boards. Increasing committee independence does not increase firm value, but in no case do we see a decrease in firm value. Economists have debated whether regulations benefit or harm firms, and our work adds to the debate. If the impetus behind board activity were simply the need to comply with regulation and the fear of stockholder litigation, we would expect increases in board activity to have a negative impact on firm value, as the increased activity would detract management from focusing on running the firm. Alternatively, the regulations could have a salutary effect if the promulgated rules shift some of the bargaining power from entrenched management to the shareholders. Our results suggest that the 2002 Sarbanes-Oxley Act has led to an increase in the level of board monitoring and has in turn enhanced firm value. Acknowledgements We thank Simi Kedia, Jeffrey Netter (the editor), Darius Palia, Oded Palmon, seminar participants at Villanova University and Fordham University, participants at the 2008 Financial Management Association Meetings and the 2008 Southern Finance Association Meetings, and the anonymous referee for their comments. This research was supported in part by the Whitcomb Center for Research in Financial Services, Rutgers University and by a Faculty Research Grant from Fordham University.

552

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

Appendix A

Variable Endogenous Variables LogMEETINGS LogMONITOR Tobin Q

ROA PPS

Instrumental Variables DUMDIRPPS DIRMTGFEE RETURNM and RETURN2M

TENURE CEOAGE LAGQ

Control Variables ACQDUM TARGETDUM RESTATEDUM CEO_CHAIR BSIZE INDEP LogASSETS VOLATILITY LEVERAGE R&D G-INDEX CEOINNOM

Definition

Purpose

Log of the annual number of board meetings Log of the product of the number of independent directors and the number of annual board meetings. Ratio of Total market value of the firm (market value of Equity plus Book Value of total debt) to Total Assets. Industry adjusted using equally weighted average at the two-digit SIC level. The ratio of EBITDA to Total Assets. Industry adjusted using equally weighted average at the two-digit SIC level. PPS is the dollar value changes in the CEO's stock and options portfolio for a one percent change in the aggregate value of the firm equity.

A monitoring proxy A monitoring proxy

Unitary variable that equals one if director compensation contains an equity component. The fee the director receives per meeting attended (in $thousands). RETURNM (RETURN2M) is the one-year holding period stock return for the prior (two-years prior) fiscal year. Industry adjusted using equally weighted average at the two-digit SIC level. Length of time that the CEO has been on the job. The age of the CEO. Lagged Tobin Q

A unitary variable that equals to one if the firm acquired another firm during that fiscal year. A unitary variable that is equal to one if the firm was a target for a merger bid during the fiscal year. A unitary variable that is equal to one if the firm restated its earnings A unitary variable if the CEO of the firm is also the Chair of the Board of Directors The number of directors on the board. The fraction of independent directors Logarithm of the level of Total Assets. Annualized standard deviation of monthly stock returns for the 60 months preceding the end of the fiscal year. Long-term debt/ Total Assets. Level of R&D expenses scaled by Total Assets. Gompers, et al. (2003) governance index. An unitary variable that is equal to one if the CEO is a member of the Nominating Committee

Firm value proxy

Operating performance proxy Pay-for-performance sensitivity proxy

Instrument for monitoring. We expect a positive coefficient. Instrument for monitoring. We expect a positive coefficient. Instrument for monitoring. We expect a negative coefficient. Instrumental for PPS. We expect a positive coefficient. Instrumental for PPS. We expect a positive coefficient. Instrumental for Tobin Q. We expect a positive coefficient.

Positive coefficient Positive coefficient Positive coefficient No prediction Positive coefficient Positive coefficient Positive coefficient No prediction No No No No

prediction prediction prediction prediction

References Adams, R., 2005. What do boards do? Evidence from board committee and director compensation data. Working Paper. Stockholm School of Economics. Blue Ribbon Committee on Improving the Effectiveness of Corporate Audit Committees, 1999. Report and Recommendation. NYSE and Nasdaq. Boone, A., Field, L., Karpoff, J., Raheja, C., 2007. The determinants of corporate board size and composition: An empirical analysis. J. Financ. Econ. 85, 66–101. Brick, I., Chidambaran, N.K., 2008. Board monitoring, firm risk and external regulation. J. Regul. Econ. 33, 87–116. Brick, I., Palia, D., Chia, J.W., 2008. Simultaneous equation of CEO compensation, leverage, and board characteristics on firm value. Working Paper. Rutgers University. Burns, N., Kedia, S., 2006. The impact of CEO incentives on misreporting. J. Financ. Econ. 79, 35–67. Byrd, J., Hickman, K., 1992. Do outside directors monitor managers? Evidence from tender offer bids. J. Financ. Econ. 32, 195–221. Chhaochharia, V., Grinstein, Y., 2007. Corporate governance and firm value: The impact of the 2002 governance rules. J. Finance 52, 1789–1825. Coles, J.L., Daniel, N.D., Naveen, L., 2006. Managerial incentives and risk-taking. J. Financ. Econ. 79, 431–468. Coles, J.L., Daniel, N.D., Naveen, L., 2008. Boards: Does one size fit all? J. Financ. Econ. 87, 329–356. Core, J.L., Guay, W., 2002. Estimating the value of employee stock option portfolios and their sensitivities to price and volatility. J. Acc. Res. 40, 613–630. Cotter, J.F., Shivdasani, A., Zenner, M., 1997. Do independent directors enhance target shareholder wealth during tender offers? J. Financ. Econ. 43, 195–218. Davidson, R., MacKinnon, J.G., 2004. Estimation and Inference in Econometrics. Oxford University Press, New York. Demsetz, H., Villalonga, B., 2001. Ownership structure and corporate performance. J. Corp. Finance 7, 209–233. Dennis, D., Hanouna, P., Sarin, A., 2006. Is there a dark side to incentive compensation? J. Corp. Finance 12, 467–488. Gibbons, R., Murphy, K., 1992. Optimal incentives contracts in the presence of career concerns: theory and evidence. J. Polit. Econ. 88, 468–505. Gompers, P., Ishii, J., Metrick, A., 2003. Corporate governance and equity prices. Q. J. Econ. 118, 107–155. Hermalin, B., Weisbach, M., 2006. A framework for assessing corporate governance reform. NBER Working Paper No. W12050. Jensen, M., 1993. The modern industrial revolution, exit, and the failure of internal control systems. J. Finance 831–880.

I.E. Brick, N.K. Chidambaran / Journal of Corporate Finance 16 (2010) 533–553

553

Karamanou, I., Vafeas, N., 2005. The association between corporate boards, audit committees, and management earnings forecasts: An empirical analysis. J. Acc. Res. 43, 453–486. Kole, S., 1995. Measuring managerial equity ownership: a comparison of sources of ownership data. J. Corp. Finance 1, 413–435. Klein, A., 2002. Audit committee, board of director characteristics, and earnings management. J. Accounting and Economics 33, 375–400. Lambert, R., Larcker, D., Verrecchia, R., 1991. Portfolio consideration in valuing executive compensation. J. Acc. Res. 29, 129–149. Linck, J., Netter, J., Yang, T., 2008. The determinants of board structure. J. Financ. Econ. 87, 308–328. Mace, M., 1986. Directors: Myth and Reality. Harvard Business School Press, Boston, MA. McConnell, J., Servaes, H., 1990. Additional evidence on equity ownership and corporate value. J. Financ. Econ. 27, 595–612. Palia, D., 2001. Endogeneity of managerial compensation in firm valuation: A solution. Rev. Financ. Stud. 14, 735–764. Prendergast, C., 2000. What trade-off of risk and incentives? Am. Econ. Rev. 90, 421–425. Raheja, C., 2005. Determinants of board size and composition: A theory of corporate boards. J. Financ. Quant. Anal. 40, 283–306. Reeb, D., Upadhyay, A., 2010. Subordinate board structures. Journal of Corporate Finance, Forthcoming Vafeas, N., 1999. Board meeting frequency and firm performance. J. Financ. Econ. 53, 113–142. Weisbach, M., 1988. Outside directors and CEO turnover. J. Financ. Econ. 20, 431–460. Wintoki, M.B., Linck, J.S., Netter, J.M., 2010. Endogeneity and the Dynamics of Internal Corporate Governance. Working Paper. University of Georgia. Yermack, D., 1996. Higher market valuation of companies with a small board of directors. J. Financ. Econ. 40, 185–221.