Are interlocked directors effective monitors?
Jamie Dimon, President and Chief Operating Officer of J.P. Morgan Chase & Co., stepped down from the board of fast-food restaurant operator Yum! Brands Inc. last year amid concerns of potential conflicts from interlocking directorships. David C. Novak, Yum! Brands Chairman and Chief Executive, sits on J.P. Morgan's board. '7 was on his board and he was on my board and we just said let's eliminate this conflict," says Mr. Dimon (Raghavan, 2005).
Financial economists have long recognized that the separation of ownership and control in large US corporations creates the potential for costly agency conflicts. Dispersed shareholders have limited incentive to monitor agents hired to run their firm that allows managers to indulge in self-serving behavior at the expense of shareholders' wealth. The board of directors, in principle, provides a mechanism for dispersed shareholders to monitor managers and ensure that major corporate actions and the setting of compensation for top level managers are undertaken in the interests of shareholders.
An extensive literature has examined how the effectiveness of monitoring by the board of directors can impact firm performance and corporate decision making. A range of board characteristics hypothesized to be correlated to board effectiveness has been examined including board size (Yermack, 1996), proportion of outside directors (Cotter, Shivdasani, and Zenner, 1997), proportion of outside directors over age 69 years (Core, Holthausen, and Larcker, 1999), equity and option ownership by outside directors (Byrd and Hickman, 1992), structure of compensation and audit committees (Klein, 1998), frequency of board and committee meetings (Vafeas, 1999), classified boards (Faleye, 2007), and number of external appointments held by directors (Ferris, Jagannathan, and Pritchard, 2003).
A more recent stream in this line of research suggests that the presence of interlocked directors and connected boards may compromise the effectiveness of board monitoring, especially with respect to the setting of compensation of CEOs. Hallock (1997), Larcker, Richardson, Seary, and Tuna (2006), and Barnea and Guedj (2006) find that interlocked and connected boards result in higher levels of CEO compensation, after controlling for economic determinants, board structure, and CEO characteristics. Other research also indicates that interlocked boards are associated with self-serving behavior by CEOs in the areas of accounting discretion and financial accounting fraud (Erickson, Hanlon, and Maydew, 2006; Bowen, Rajgopal, and Venkatachalam, 2008). (1)
Our paper extends the literature on interrelated boards by providing a detailed examination of whether the presence of interlocked directors on a board is associated with weak governance. Specifically, we examine whether the presence of interlocked directors is associated with lower industry-adjusted firm performance. We also study the stock price reaction to announcements of director appointments that create interlocked directors to test how shareholders view the presence of interlocked directors. To provide further evidence regarding the association between interlocked directors and board effectiveness, we observe the effect of interlocked directors on the pay-performance sensitivity of CEO incentive compensation and on the sensitivity of CEO turnover to prior firm performance.
We hypothesize that the excess compensation levels of interlocked firms and connected firms documented by the earlier cited papers may be indicative of weak governance and entrenched management. Poorly governed firms are likely to have lower firm performance. Since the incidence of interlocked directors and firm performance may be jointly determined by the underlying quality of corporate governance, we employ a simultaneous equations framework to control for potential endogeneity between the presence of interlocks and firm performance. This allows us to properly examine the relationship between firm performance and the presence of interlocked directors.
For a sample of 3,566 firm-years over the 2001-2003 time period, we find that poorly performing firms are more likely to have interlocked directors on their boards. We obtain this result using either industry-adjusted Tobin's Q or industry-adjusted ROA as the firm performance measure. We also find that the presence of interlocked directors negatively impacts industry-adjusted ROA, but this result is not always robust to the use of industry-adjusted Tobin's Q as the performance measure.
In further analysis, we find that the market reacts negatively to the announcement of director appointments that create interlocked boards. In addition, we find that interlocked directors are associated with lower than optimal pay-performance sensitivity of CEO incentive compensation. Finally, we find evidence that interlocked directors lower the sensitivity of CEO turnover to prior firm performance. Collectively, our results suggest that the presence of interlocked directors is indicative of poorly governed firms. From a public policy perspective, regulatory authorities and activist institutional investors may take these findings into consideration when making recommendations on the optimal structure of corporate boards.
The remainder of the paper is organized as follows. The next section surveys related literature on interlocked and connected boards and provides the motivation for our paper. Section II provides details regarding data issues and empirical methodology. Section III discusses our empirical results, and our conclusions are presented in Section IV.
I. Literature Review and Motivation
Earlier work by organizational theorists broadly proposed two opposing viewpoints as to why interlocked boards may arise. Pfeffer and Salancik (1978) and Bazerman and Schoorman (1983) propose a resource dependency theory that contends that interorganizational linkages, such as those linked through interlocked directors, serve to buffer the effects of environmental uncertainty. This viewpoint is supported by findings in Mizruchi and Stearns (1988) and Richardson (1987). In contrast, class integration theories (Koenig and Gogel, 1981) posit that interlocked directorates arise in order to help protect the interests of members of a social class and, as such, should have an inconsequential or potentially negative impact on firm performance. This view is supported by findings in Useem (1982) and Nguyen-Dang (2007).
A recent stream of literature examines the determinants of interlocked boards and the monitoring effectiveness of boards with directors holding multiple directorships. Fich and White (2005) find that the likelihood of CEO interlocks is higher for CEOs of firms whose board members hold higher numbers of directorships. They argue that this is consistent with CEOs of high-quality boards (proxied by the number of directorships held by its board members) who are more likely to be sought after as board members for other firms. An alternative interpretation is that firms with boards comprising "busy" directors may indicate entrenched management since "busy" directors may not perform their monitoring role effectively. The evidence on the effects of "busy" directors is mixed. Ferris, Jagannathan, and Pritchard (2003) find no evidence that multiple directorships have a deleterious impact on important board functions or subsequent firm performance, while Fich and Shivdasani (2006) find that boards in which a majority of outside directors hold three or more directorships are associated with weak governance.
Recent research also examines the association between interlocked boards and CEO compensation. Hallock (1997) interprets his results as indicating that interlocked boards, particularly those that are CEO interlocked, contribute to higher CEO pay. Using a broader definition of director connectedness called "back-door distance," Larcker et al. (2006) document that CEOs at firms where there is relatively short back-door distance between inside and outside directors or between the CEO and members of the compensation committee earn substantially higher levels of total compensation after controlling for economic determinants, board structure, and CEO characteristics. Barnea and Guedj (2006) use an even broader definition of connectedness to demonstrate how networks of directors affect CEO compensation. They find that CEOs of firms that have more connected board members have higher levels of CEO compensation.
Other papers have used the number of interlocked outside directors as a proxy for the effectiveness of board oversight in different contexts. Bowen, Rajgopal, and Venkatachalam (2008) find that firms with more interlocked directors on the board exercise greater accounting discretion, potentially contributing to excess CEO compensation. Similarly, Erickson, Hanlon, and Maydew (2006) illustrate that the compensation-based incentive to commit financial accounting fraud is significantly positively related to the occurrence of interlocked boards and other variables that are suggestive of less board independence. Overall, the above research suggests that interlocked directors may not perform their monitoring role effectively.
II. Data and Empirical Methodology
Our data on firms and directors are obtained primarily from the Corporate Library for the 2001-2003 proxy seasons. (2) The Corporate Library data set provides detailed information on 5,302 firms and their 58,153 directors. We exclude firms listed without share codes 10 or 11 (ordinary shares from the Center for Research in Security Prices [CRSP]), foreign firms, and firms that are subsidiaries. We also exclude 2,187 director observations where the director status is not coded as "director" (i.e., where the individual is deceased, retired, and otherwise inactive). Following previous work (Hallock, 1997), we focus on interlocked directors that involve one inside director and one outside director since these types of interlocks are the most likely to compromise the monitoring effectiveness of boards. (3) We classify directors as "inside" if they are labeled "inside" by the Corporate Library, while directors are considered "outside" if labeled by the Corporate Library as "unrelated" or "related" outside directors. Specifically, we define an interlock as occurring when Firm A has Director 1 serving as an insider and Director 2 serving as an outsider, whereas Firm B has Director 1 serving as an outsider and Director 2 serving as an insider.
Merging with the Compustat database reduces the sample by 688 observations. Because we wish to compare the impact of interlocked boards as defined above to noninterlocked boards, we eliminate 330 firm-years involving other types of interlock (e.g., those involving two outside directors). Finally, 397 firm-years have missing Compustat control variables. This brings the final sample to 3,566 firm-year observations comprising 1,463 unique firms. Of these 3,566 firm-years, a total of 118 firm-years have one or more interlocked directors. Specific details on how we obtain the final sample are provided in Table I. This sample is used throughout the ensuing analysis.
We next examine, in greater detail, the interlocked observations in our sample period. The 118 interlocked firm-years involve 139 pairs of interlocked directors with 71 occurring in 2001, 39 in 2002, and 29 in 2003, as reported in Table II, Panel A. Thus, the incidence of interlocked directors declines following the passage of the Sarbanes-Oxley Act. In the new corporate governance regime involving higher levels of information disclosure, firms may decide to restructure their boards to eliminate explicit director interrelationships that may be viewed negatively by shareholders, especially activist institutional investors.
Since any pair of interlocked directors may repeat from year to year if their directorships do not change, we also identify their initial occurrence in our sample period. In Panel B, we illustrate that there are 94 unique occurrences over the sample period with 71 in 2001, 11 in 2002, and 12 in 2003. Panel C describes various characteristics of these 94 interlocked director pairs. About 12% of the 94 interlocks occur within the same (one-digit) Standard Industrial Classification (SIC) code. Interlocked financial firms make up 32% of the sample, and 33% of the interlocks are firms that are interlocked with financials. In 32 cases (34%), the insider involved in the interlock is a CEO while the outsider involved in the interlock is a CEO in 37 of the interlocks (39.4%).
B. Empirical Methodology for Examining the Interlock-Firm Performance Relationship
To provide evidence regarding whether interlocked directors are associated with weak monitoring, we examine the association between the presence of interlocked directors and firm performance. Since the incidence of interlocked directors and firm performance may be jointly determined by the underlying quality of corporate governance, we employ a simultaneous equations framework to control for potential endogeneity between the presence of interlocks and firm performance. This allows us to correctly examine the relationship between firm performance and the presence of interlocked directors (Greene, 2000).
We use a two-step instrumental variables (IV) methodology. The coefficient estimates are generated using the two-stage probit least squares (2SPLS) approach (Maddala, 1983) designed for models with one discrete and one continuous dependent variable. In the 2SPLS approach, reduced-form equations for the endogenous variables are first estimated using ordinary least squares for the continuous variable (firm performance) and a probit model for the binary variable (occurrence of interlocks). A predicted value for each endogenous variable is generated using the estimates from the reduced-form equations. The equations are estimated in the second stage using these predicted values in place of the endogenous variables on the right-hand side of the respective equations. The estimates from this second stage have been shown to be consistent (Amemiya, 1978; Achen, 1986). We adjust the coefficients' standard errors in the second stage to correct for lack of independence in the errors arising from the nature of panel data (i.e., the standard errors are clustered at the firm level).
We model performance as a function of interlock (the endogenous regressor), a series of exogenous variables that also affect interlock, and additional exogenous variables that are presumed to only affect performance. The performance equation takes the following form
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (1)
Similarly, we model interlock as a function of performance, a set of independent variables that are also related to performance, and additional exogenous variables that only affect interlock. The interlock equation is estimated as follows:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (2)
The first endogenous variable, Performance, is measured by industry-adjusted Tobin's Q and, alternatively, industry-adjusted ROA. We follow standard practice in the literature by using Q = (MVE - BVE + TA)/TA as a proxy for the theoretical Q, where MVE is the market value of equity, BVE is the book value of equity, and TA is the book value of the firm's total assets. Q and ROA are adjusted using the median of all Compustat companies at the two-digit SIC level. The second endogenous variable, Interlock, equals one if a given board has one or more director interlocks with other boards according to our definition of interlock discussed above and 0 otherwise. We next provide details of the exogenous variables used in the two equations and their predicted signs.
1. Firm Performance Equation
The exogenous variables of the firm performance equation include alternative mechanisms that control agency problems and firm-specific characteristics that are likely to explain Tobin's Q. A number of the variables reflect measures of board independence and ownership structure. With respect to the proportion of inside directors on the board (Inside directors), we do not have a prior expectation of the sign. While greater board independence should ostensibly be associated with higher levels of performance, the literature addressing this relation finds positive, negative, and insignificant associations that appear to be conditional based upon how the relationship is estimated and how performance is measured (John and Senbet, 1998). Similar to Core, Holthausen, and Larcker (1999), we include the proportion of directors that have a current financial relationship with the firm or are former employees as an additional measure of independence (Gray directors). In a similar vein, the board structure literature suggests that boards chaired by the CEO (CEO duality) may be less independent due to the decision-making authority of the board chairman (Shivdasani and Yermack, 1999). However, the empirical relation between CEO duality and performance remains unclear (Baliga, Moyer, and Rao, 1996). The log of Board size is expected to be negatively related to performance since Yermack (1996) provides evidence that smaller boards are associated with higher levels of Q, supporting the view that decision making is less efficient for larger boards. Outside over 3 boards is calculated as the number of outside directors holding four or more directorships divided by the total number of outside directors. This is an important control variable that enables us to distinguish between the effects of interlocked directors and "busy" directors. Fich and Shivdasani (2006) find that "busy" boards are associated with weak governance while Ferris, Jagannathan, and Pritehard (2003) do not find any negative effects associated with "busy" directors. Fama and Jensen (1983) also argue that multiple directorships are a signal of reputational capital. Outside over 69 is calculated as the number of outside directors over the age of 69 years divided by the total number of outside directors. Core, Holthausen, and Larcker (1999) posit that outside directors become less effective as they become older or serve on too many boards, and find that greater proportions of each measure are associated with higher levels of CEO compensation. Classified board is a binary variable that takes the value of one if the board is classified (i.e., directors stand for reelection on a staggered schedule) and zero otherwise. We expect Classified board to be negatively related to Q since Bebchuk and Cohen (2005) and Faleye (2007) find that classified boards are associated with entrenched management and diminished performance. With respect to ownership structure, Inside ownership is the proportion of equity held by top management and directors and gauges the degree of alignment between shareholder and managerial interests. We control for any non-linearities in the ownership-performance relation (McConnell and Servaes, 1990) by including the square of Inside ownership.
The remaining independent variables control for financial and other firm characteristics. We include the binary variable Governance policy to test whether firms with a formal corporate governance policy perform better than those without formal policies. Three-year sales CAGR is the compounded annual growth rate of sales over the prior three years and controls for future growth opportunities. We expect Long-term debt (book value of long-term debt divided by total assets) to be positively related to firm performance due to its use as a disciplining mechanism (Jensen, 1986) and as a corporate tax shield (Modigliani and Miller, 1963). The log of Sales controls for differences in firm size. The previous literature documents that stock performance is inversely related to size (Fama and French, 1992). Finally, we include Y02 and Y03, each of which equals one if the year is 2002 and 2003, respectively, to control for any secular changes in performance over the sample period.
2. Interlock Equation
The second equation in the system models the binary variable Interlock. Since director interlocks have generally been viewed in past empirical work as suggestive of weak monitoring, we select board structure variables that have been associated with the effectiveness of monitoring. These characteristics also control for factors that may mechanically influence the likelihood of the formation of interlocks. The log of the total number of directors (Board size) is expected to be positively related to the likelihood of an interlock since board size is inversely associated with better monitoring (Yermack, 1996). The predicted sign of Outside over 3 boards is unclear since there is conflicting evidence of its effect on monitoring (Ferris, Jagannathan, and Pritchard, 2003; Fich and Shivdasani, 2006). We expect Outside over 69 to be positively related to Interlock if this characteristic is indicative of a weak board.
We include several variables that proxy for CEO influence and alignment with shareholder interests. CEO age measures the age of the current CEO while CEO tenure measures the number of years that the current CEO has been in that position. Long-serving or older CEOs may be more likely to seek outside opportunities on other boards or be more likely to tolerate or encourage other corporate insiders to seek outside directorships. CEO ownership is defined as the percentage of equity held by the CEO and proxies for how tightly the CEO's interests are aligned with their firm's performance (Jensen and Meckling, 1976). To the extent that lower proportions of CEO equity ownership reduce the incentive to devote attention to their own board, we expect lower levels of CEO ownership to be positively correlated with the incidence of interlocked boards. Institutional ownership tests whether greater proportions of equity ownership by institutions discourage interlocks. Finally, the binary variable Governance policy controls for whether an explicit corporate governance policy has any bearing on the incentive for insiders to interlock with outside boards.
The remaining control variables include the log of Sales that is expected to be positively related to Interlock if the visibility associated with larger firms affords the opportunity for its directors to obtain directorships on other boards. Also included are FIN, a binary variable that takes the value of one if the firm is in the financial sector (SIC code 6000-6999) and zero otherwise. The sign and significance of this variable tests the resource dependency theory, which posits that nonfinancial firms may be more likely to interlock with financials as a way of securing lower cost capital (Richardson, 1987). We include Y02 and Y03, each of which equals one if the year is 2002 and 2003, respectively, to control for any secular changes in the likelihood of interlocks.
III. Empirical Analysis
A. Descriptive Statistics and Univariate Analysis
Table III, Panels A-D provide aggregate descriptive statistics of the endogenous and exogenous variables used in our performance analysis. We provide descriptions of the variables in the Appendix. Panel A reports statistics for the endogenous variables consisting of the incidence of interlocks and performance measures (unadjusted and industry-adjusted Q and, alternately, unadjusted and industry-adjusted ROA). Interlocked firms comprise 3.3% of the sample. In comparison, Fich and White (2005) report that approximately 14.3% of 610 firms in 1991 are CEO interlocked, indicating that the proportion of interlocked boards has declined significantly since then. We also report in Table II a similar declining trend over our sample period 2001-2003. The mean (median) value of Q is 1.933 (1.403) while the mean (median) value of ROA is 1.7% (3.5%). The mean (median) value of industry-adjusted Q is 0.545 (0.100) while the mean (median) value of industry-adjusted ROA is 4.0% (3.4%). Because of the relative importance of industry specific factors for firm performance, we focus our analysis on industry-adjusted Q and ROA.
Panel B provides descriptive statistics of variables pertaining to board composition and structure. The mean (median) proportion of corporate insiders on the board is 22.8% (18.2%), while gray directors comprise an average of 8.3% of the board. On average, 69.7% of all boards have chairs that also hold the position of CEO. The mean (median) board size in our sample is 9.35 (9). The proportion of outside directors (relative to all outside directors) that sit on four or more boards is 13.8% while the proportion of outsiders who are greater than 69 years old is 10.6%. Finally, about 60% of all firms have classified boards.
Turning to CEO characteristics and ownership structure in Panel C, the mean (median) CEO age is 54.8 (55) years old and the mean (median) CEO tenure is 7.8 (5) years, similar to the 8.62 years reported by Fich and White (2005). The mean (median) CEO share ownership is 2.7% (0.3%) while the mean (median) ownership of corporate insiders collectively is 22.0% (18%). Finally, the mean (median) institutional ownership is 61.9% (63.8%).
The remaining financial control variables are given in Panel D. A minority of firms (12.5%) have an explicit statement regarding corporate governance. The mean (median) compounded annual growth rate of sales over the three years prior to each corresponding sample year is approximately 10.7% (7.8%). The median firm's capital structure comprises 15.7% long-term debt, and about 14.3% of the firms in our sample are in the financial sector. Mean (median) firm size measured by total sales is $4,507 million ($1,110 million).
Table IV reports the means and medians of performance measures, governance, and financial control variables for the 118 firm-years of interlocked firms (Column 2) and the 3,448 firm-years of noninterlocked firms in the sample (Column 3), together with the p-values for differences in means (medians) using the standard difference-in-means test (Wilcoxon test) in Column 4. We highlight characteristics that have statistically different means and medians between the interlocked and noninterlocked samples (Columns 2-3). Although the median unadjusted Q is significantly lower for interlocked firms, there is no statistical significance when using industry-adjusted Q. On the other hand, the median industry-adjusted ROA is significantly lower for interlocked firms. In Panel B, interlocked firms have significantly lower mean and median proportions of insiders and gray directors on the board and have fewer incidences of CEOs who also hold the position of chairman of the board. Boards of interlocked firms are significantly larger as compared to those of non-interlocked firms, and have significantly greater proportions of outside directors that sit on four or more boards. Interlocked firms tend to have lower proportions of outside directors over the age of 69 years. Finally, classified boards are more prevalent among interlocked firms. Panel C illustrates that the mean (median) age of the CEO is higher for interlocked firms and the mean (median) institutional ownership is lower for interlocked firms. Among the remaining control variables in Panel D, we find that interlocked firms are more likely to have an explicit statement on corporate governance, have lower prior three-year sales growth rates, have higher levels of long-term debt, are more likely to be in the financial sector, and are significantly larger than non-interlocked firms.
One characteristic that is significantly different between the interlocked and noninterlocked samples is firm size. Board structure variables, in particular, can be correlated with firm size; for example, larger firms tend to have larger boards. To investigate whether differences in firm size are driving some of the above results, we examine differences in characteristics between interlocked and noninterlocked firms after restricting the noninterlocked control firms to those with Sales above the median ($1,087 million) of the original sample (1,724 firm-years out of the original 3,448). The resulting mean (median) firm size of the size-constrained control sample is $8,046 ($3,031) million that is more comparable to $11,568 ($3,804) million of the interlocked sample. Table IV compares performance measures, governance, and financial control variables for the 118 firm-years of interlocked firms (Column 2) to the size-constrained 1,724 firm-years of noninterloeked firms (Column 5). The p-values for differences in means (medians) are given in Column 6. The univariate comparisons indicate that while the magnitude of the differences tends to diminish, the results are, for the most part, similar to those using the full noninterlocked sample.
B. Multivariate Analysis of Interlocked Directors and Firm Performance
Our next step is to examine the association of incidence of interlocks and firm performance using the 2SPLS methodology described in Section III.B. Table V provides coefficient estimates of the system of equations using firm performance and Interlock as endogenous variables using the full sample. Model 1 uses industry-adjusted Tobin's Q as the performance measure and Model 2 uses industry-adjusted ROA as the alternative performance measure.
We first test for endogeneity of firm performance and incidence of interlocks. We use the Durbin-Wu-Hausman and Wald exogeneity tests for the performance and interlock equations, respectively. (4) The p-values of the Durbin-Wu-Hausman test (0.000 for industry-adjusted Q and 0.024 for industry-adjusted ROA) and Wald test (0.001 in Model 1 and 0.058 in Model 2 for interlock) suggest that firm performance and incidence of interlocks are indeed endogenously related. We attempt to identify the system of equations with powerful and valid instruments. We use the prior literature and existing theory to identify valid instruments for each equation as discussed in Sections III.B. 1 and III.B.2. We then check to see if these instruments are significantly related to each endogenous dependent variable. The first-stage regression with industry-adjusted Q (industry-adjusted ROA) has an [R.sup.2] of 0.15 (0.11) while the interlock equation has a pseudo R2 of 0.15. This indicates that the instruments are significantly related to each endogenous variable. Finally, we use the Hansen J test (performance equation) and the Amemiya-Lee-Newey test (interlock equation) to ensure that the system of equations is well identified. The p-value of the Hansen J statistic (0.158 for industry-adjusted Q and 0.19 for industry-adjusted ROA) and the p-value of the Amemiya-Lee-Newey statistic (0.383 in Model 1 and 0.130 in Model 2 for interlock) suggest that the instruments used are valid and that the system of equations is well identified.
In Model 1, the first equation demonstrates that the coefficient of the endogenous Interlock variable, while negative in sign, is not significant at the 10% significance level. Thus, the presence of interlocked directors does not appear to significantly lower firm value as measured by industry-adjusted Tobin's Q. Among the variables representing board independence and ownership structure, we find that the proportion of inside directors is positively related to industry-adjusted Q at the 5% level. This result is consistent with the findings of Agrawal and Knoeber (1996) that greater proportions of outsiders negatively affect performance. The coefficient on the proportion of outside directors who serve on four or more boards is positive and significant. This is consistent with multiple directorships as a signal of reputation and quality (Fama and Jensen, 1983; Ferris, Jagannathan, and Pritchard, 2003; Fich and White, 2005). The coefficient of prior three-year sales growth rate is positive and significant, indicating that faster prior growth suggests higher future growth prospects that appears as a higher industry-adjusted Q. Long-term debt is negative at the 1% level, suggesting that higher proportions of long-term debt are indicative of higher financial distress risk.
In the second equation of Model 1 using Interlock as the dependent variable, the coefficient on industry-adjusted Q is negative and significant at the 1% level. This suggests that poorly performing firms are more likely to be interlocked. With respect to the board and ownership structure variables in the Interlock equation, we find that Board size is positive and significant at the 1% level. This may indicate that boards that are less effective are more likely to be interlocked since Yermack (1996) indicates that smaller boards are associated with higher levels of Tobin's Q. This may also have a mechanical explanation since the larger the board, the more opportunities for interlocks to occur. The coefficient on the proportion of outside directors that sit on four or more boards is positive and significant at the 1% level. This may indicate that boards that are less effective are more likely to be interlocked since Fich and Shivdasani (2006) find that "busy" boards are associated with weak governance. This may also have a mechanical explanation since the more outside directors on a board with several directorships, the more opportunities for insiders to interlock with them.
The coefficient of the institutional ownership variable (Institutional ownership) is negative and significant at the 5% level, indicating that firms with higher levels of institutional equity ownership have lower incidence of interlocks. The coefficient on log (Sales) is positive and significant, suggesting that the visibility associated with larger firms may create opportunities for its directors to obtain directorships on other boards, thus increasing the likelihood of interlock. Finally, the coefficients of the year dummies demonstrate that the frequency of interlocks is declining over our sample period.
In Model 2, we reestimate the system of equations using industry-adjusted ROA in place of industry-adjusted Q as the performance measure. The results are reported in Table V, Model 2. Consistent with findings in a contemporary working paper by Larcker et al. (2006), we find that the presence of interlocked directors negatively impacts industry-adjusted ROA and this result is significant at the 1% level. Hence, we find that that the presence of interlocked directors adversely impacts firm performance using an accounting measure of firm performance. There are some differences in the coefficients of some control variables in the performance equation when using industry-adjusted ROA in place of industry-adjusted Q. This is to be expected since ROA is an accounting measure of profitability while Q captures the market's expectation of growth opportunities. In the second equation of Model 2 using Interlock as the dependent variable, the coefficient of industry-adjusted ROA is negative and significant at the 5% level. Thus, using industry-adjusted ROA as performance measure, we find that poorly performing firms are more likely to be interlocked, consistent with results obtained using industry-adjusted Tobin's Q. The coefficients of the other variables in the Interlock equation are qualitatively similar in Models 1 and 2.
As discussed earlier, our measure of whether firms have interlocked directors is a dummy variable that equals one when there is at least one interlocked director. Since a greater proportion of directors that are interlocked may indicate even more severe problems of governance and entrenchment, as a robustness test, we create an alternate measure of interlocked directors calculated as the number of interlocked directors as a proportion of the total number of directors. In unreported results, using instrumental variable regressions for the performance equation and instrumental variable tobit regressions for the interlock equation with the new interlock measure, we find results similar to that obtained using the interlock dummy variable.
We also repeat the 2SPLS analysis using the size-constrained control sample described above. The main results are similar to that obtained using the full sample. Poorly performing firms are more likely to be interlocked using both industry-adjusted Q and industry-adjusted ROA as performance measures. Also, the presence of interlocked directors negatively impacts industry-adjusted ROA and industry-adjusted Q, but the statistical significance is stronger when using industry-adjusted ROA (1% level as compared to a 10% level using industry-adjusted Q).
In order to ascertain whether performance differences are driven by poor prior performance among interlocked firms, we conduct univariate pairwise analyses using matched firms based on prior performance, industry, and size during the 2001-2003 sample period. We select matching firms based on prior performance (one-year lagged ROA), size (total assets), and industry (two-digit SIC). (5) For each interlocked firm year in our sample period 2001-2003, we choose matched firms with one-year lagged ROA within [+ or -]10% and total assets within [+ or -]30% of that of the interlocked firm. From these, we choose the best match (closest in prior performance) in the same two-digit SIC. If no matching firm is found, we relax the industry constraint to one-digit SIC. In the event we still do not find a match, we match only on preperformance and size. In terms of quality of matches, there are no significant differences in prior performance between interlocked firms and matched firms during the matching period; however, interlocked firms are significantly larger than their matched counterparts due to the relatively wide range of sizes needed to ensure that matches are obtained. Table VI, Panel A reports results using the above matching criteria. Generally, subsequent performance of interlocked firms is lower than that of matched firms. The median ROA of interlocked firms is significantly lower than that of control firms at the 10% level while the mean (median) industry-adjusted ROA is significantly lower at the 10% (5%) levels. The differences are, however, not significant using unadjusted Q and industry-adjusted Q. To verify these results in a multivariate context, we repeat the 2SPLS analysis using matching firms based on the criteria discussed above. To increase the power of the tests for the 2SPLS analysis, we choose five closest matches per sample firm. In unreported results, the coefficient estimates of the key variables are similar to those obtained using the full sample.
Finally, we extend this line of inquiry to the year the interlocking relationship was formed. As in the prior analysis, we employ the three-way matching algorithm described above. Since any pair of interlocked directors may repeat from year to year if their directorships do not change, focusing on the initial formation of interlocks results in fewer (70) observations than reported for the 2001-2003 sample period (118). Table VI, Panel B indicates that there are no statistically significant differences between the interlocked firms and control firms the year the interlock formed, using unadjusted Q, industry-adjusted Q, and industry-adjusted ROA as the performance measure. Interlocked firms have a statistically higher ROA than control firms. Viewed collectively, the multivariate and univariate analyses suggest that the negative association between interlocked directors and firm performance is driven primarily by poorly performing firms more likely to be interlocked.
C. Wealth Effects Associated with the Formation of Interlocks
Several studies find that the unconditional stock price reaction to outside director appointments is positive (Rosenstein and Wyatt, 1990; Shivdasani and Yermack, 1999). We follow this line of research by examining the stock price reaction to announcements of director appointments that create interlocked directors. The creation of interlocked directors typically arises in a sequential manner; for example, an inside director at Firm A agrees to serve on the board of its outside director (who is an inside director at Firm B), creating an interlocked directorship. However, no change in the composition of the board of Firm A takes place except that the appointment of its inside director to the board of Firm B now makes the relationship between two of its existing directors interlocked. Note that there is a change in the composition of the board of Firm B. The sample of firms that may best isolate how shareholders view interlocked directors appears to be firms like Firm A where two directors are already on their board but become interlocked when there are changes on another firm's (Firm B) board. We expect the cleanest result from this sample since no contaminating effects of board changes occur as would happen in a sample of firms like Firm B. For instance, in the case of Firm B, an appointment of an outside director creates an interlock, but any negative effect associated with interlocked directors may be offset by the positive reaction typically found for outside director appointments.
We search the Factiva database for business press announcements that create interlocked directors (i.e., the announcement date of the last director appointment that creates the director interlock). We find announcement dates for director appointments that create 69 interlocked directors, with the earliest date in 1987 and the most recent date in 2002. In Table VII, we report the stock price reaction to these announcements for the full sample and various subsamples, using standard event study methodology. Since it is not clear how quickly stock prices incorporate information revealed by these announcements, we report equally weighted and median cumulative abnormal returns (CARs) for the (- 1, + 1) and (- 1, + 10) windows surrounding the announcement date.
The market model parameters are estimated using the CRSP equal-weighted index over an estimation period of 221 days to 21 days prior to the announcement date, using Scholes-Williams betas. To test whether the mean CAR for the sample is different from zero, we use the standardized cross-sectional z-statistic discussed by Boehmer, Musumeci, and Poulsen (1991) that takes into account possible increases in the variance of returns on the event date. We also report the nonparametric rank test statistic (Corrado, 1989) to test the robustness of the parametric z-statistics.
Table VII, Panel A reports CARs for the full sample. The (-1, +l) CAR is not significantly different from zero, but the (-1, +10) CAR is significantly negative. The mean (median) CAR (- 1, +10) is--1.67% (-1.26%) and is significant at the 5% level using the z- and rank statistics. Panel B reports the announcement effects on the stock price of firms (like Firm A as discussed earlier) where directors i and j are already on the board of the company when the interlock occurs. This is likely the cleanest test of how shareholders view director interlocks since no changes to board composition occur for these firms. Using a sample of 28 observations, the (-1, +1) mean (median) CAR is -1.58% (-1.46%) and is significant at the 1% level using the z-statistic and at the 5% level using the rank statistic. The rank statistic also indicates that the (- 1, + 10) CAR is significantly negative at the 10% level. We interpret these results to indicate that shareholders negatively view the news of formation of interlocked directors as this may indicate weak governance and entrenched management.
In Panel C, we report announcement effects on the stock price of firms when director i or j are added to the board of the company creating the interlock (like Firm B as discussed earlier). Using a sample of 41 observations, the (-1, +1) CAR is insignificant, while the (-1, +10) CAR is negative and marginally significant at the 10% level using the z-statistic.
We also create subsamples dependent upon whether director i or j is the CEO on the board of Company A or not. This allows us to examine whether the potential conflict of interest associated with interlocked directors is more acute when CEOs are involved. Panel D demonstrates that announcement effects of interlocks involving the CEO are not significant using (-1, +1) CAR, but the (-1, +10) CAR is negative and significant at the 5% level using the z-statistic. Panel E similarly indicates that the announcement effect for non-CEO interlocks is not significant using the (- 1, +1) CAR; however, the (- 1, +10) CAR is negative and significant at the 5% level using the rank test statistic. From these results, it does not appear that the announcement effects for CEO and non-CEO interlocks are significantly different.
D. Impact of Interlocked Directors on Optimal Equity Ownership Incentives
The firm performance and event study results suggest that interlocked directors are associated with entrenched management and weak monitoring. To provide additional insight on this issue, we examine whether the presence of interlocked directors is associated with appropriate levels of pay-performance sensitivity of CEO and top-level management incentive compensation in our sample. Hallock (1997) uses a definition of interlocked directors similar to ours and finds that CEOs of interlocked firms have higher levels of pay for a sample of large firms in 1992. (6) We extend this line of inquiry by examining the pay-performance sensitivity of CEO and top-level management incentive compensation for interlocked and noninterlocked boards.
Following recent work in the executive compensation literature (Coles, Daniel, and Naveen, 2006), we focus on delta as a measure of pay-for-performance sensitivity, where delta is the change in the value of CEO equity incentives for a 1% change in firm value. (7) Equity incentives include stock options (current options plus previously granted exercisable and unexercisable options) along with unrestricted and restricted stock grants. We employ the "one-year approximation" methodology described by Core and Guay (2002) to calculate proxies for option portfolio value and sensitivities using inputs supplied by the Execucomp database. The deltas for options, unrestricted, and restricted stock grants are aggregated for each CEO. Similarly, we calculate a management team delta for each firm by aggregating the portfolio equity incentives for all management team members recorded in Execucomp.
Our focus is not on the magnitude of incentive compensation delta of interlocked firms per se, but rather if delta of interlocked firms systematically deviates from optimal levels. The importance of examining "residual" deltas is highlighted by Tong (2008) who finds that both positive and negative deviations of CEO ownership (this measure includes option ownership multiplied by delta) from optimal levels reduces firm value. Core and Guay (1999) provide a theoretical and empirical framework for the determinants of CEOs' portfolio holdings of equity incentives. They specify a model for CEOs' portfolio holdings of equity incentives that includes manager, firm, and industry-specific variables as follows: (8)
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (3)
If the optimal level of CEO equity incentive compensation is determined by CEO, firm, and industry characteristics, it follows that in the absence of agency conflicts, the level of CEO payperformance sensitivity should be completely explained by economic factors and excess CEO compensation incentives should be zero across firms. Equity incentives that are below or above optimal levels should be explained by indicators of agency conflicts such as interlocked boards. We test this proposition by collecting the residuals from Equation (3) and running Equation (4) that is based on Core, Holthausen, and Larcker (1999). The independent variables used are described in the Appendix.
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (4)
Table VIII, Panel A provides means (medians) for portfolio equity delta and residual delta for the CEO and management team for the interlocked firms (Column 2) and noninterlocked firms (Column 4). (9) The first two rows of Panel A demonstrate that interlocked firms have higher median levels of equity portfolio delta for both the CEO and the overall management team. However, whether these levels are optimal, given the underlying economic fundamentals, is addressed by the next two rows. The univariate difference in mean (median) excess residual delta between the two samples is significantly lower for interlocked firms, indicating levels below optimal of equity incentive compensation.
The least-squares coefficient estimates for Equation (4) are presented in Table VIII, Panel B. The p-values reflect robust standard errors that are cluster adjusted at the firm level. Using the residual CEO and management team pay-performance sensitivity as alternative dependent variables, the coefficient on Interlock is negative and significant at least at the 5% level in Column 2. This indicates that CEOs of firms with interlocked boards have statistically lower residual delta relative to firms that are noninterlocked (i.e., interlocked CEOs are underincentivized relative to the optimum level of incentive compensation). Column 3 shows that residual management team delta is also negative but insignificant (p = 0.13), indicating that the impact of director interlocks have a greater impact on the pay-performance sensitivity of CEO compensation incentives than on the overall management team.
E. Impact of Interlocked Directors on CEO Turnover
As an additional test of whether interlocked directors are associated with weak monitoring, we examine the sensitivity of CEO turnover to firm performance in our sample. For this purpose, we create two measures of CEO turnover based on different criteria. First, we create a binary variable called HPS CEO turnover using the Huson, Parrino, and Starks (2001) algorithm that attempts to identify turnovers that are forced. (10) Given that there can be misclassification error even with the Huson, Parrino, and Starks algorithm, we also create a binary variable called All CEO turnover that captures sample firms CEO turnovers that occur for any reason. The CEO turnover binary variables take the value of one for firm-years with CEO turnovers and zero for firm-years with no CEO turnovers.
We estimate probit regressions with HPS CEO turnover and All CEO turnover as dependent variables during the 2001-2003 sample period with corrections for clustered errors. Firm performance is measured as market-adjusted prior one-year return, defined as the firm's stock return over the prior year minus the S&P 500 return over that period. We also include the interaction of Interlock and firm performance to measure whether the presence of interlocked directors weakens the sensitivity of CEO turnover to firm performance. We control for age-related CEO retirements in both models using the variable CEO age; this is an especially important explanatory variable when using All CEO turnover as the dependent variable. The remaining control variables are related to governance quality, monitoring effectiveness, and firm size. Year dummies are included to control for secular changes in turnover during the sample period. The results are reported in Table IX.
Using HPS CEO turnover as the dependent variable, we find a significant negative relation between the likelihood of turnover and firm performance consistent with Coughlan and Schmidt (1985). The coefficient of the interaction term of Interlock and firm performance is positive but not significant, suggesting that the presence of interlocked directors does not affect the sensitivity of CEO turnover to firm performance. Using the broader classification All CEO turnover as the dependent variable, the coefficient of the interaction term of Interlock and firm performance is positive and significant at the 5% level, suggesting that the presence of interlocked directors reduces the sensitivity of CEO turnover to firm performance.
This paper examines whether the presence of interlocked directors on a board is associated with weak governance. For a sample of 3,566 firm-years from 2001-2003, we find that firms with lower industry-adjusted firm performance are more likely to have interlocked directors. This result is robust to the use of industry-adjusted Tobin's Q and industry-adjusted ROA as the performance measures. Multivariate and univariate pairwise analyses produce mixed results as to whether the presence of interlocked directors has a negative effect on performance. This suggests that the negative association between interlocked directors and firm performance is driven primarily by poorly performing firms more likely to be interlocked.
We find a negative stock price reaction to the announcement of director appointments that create interlocked directors, suggesting that shareholders view the presence of interlocked directors as an indication of weak monitoring and entrenched management. We also document that the presence of interlocked directors is associated with lower than optimal pay-performance sensitivity of CEO incentive compensation, suggesting that boards with interlocked directors do not perform this critical function adequately. In addition, we find evidence that the incidence of interlocked directors lowers the sensitivity of CEO turnover to firm performance. Collectively, our results suggest that the presence of interlocked directors is indicative of weak governance and entrenched managers.
Appendix: Description of the Variables Variable Name Description and Source Panel A. Endogenous Variables Q Tobin's Q is the market value of equity minus the book value of equity plus total assets, divided by the book value of total assets. Source: Compustat. Industry adj. Q Firm Q minus the median (from the Compustat population for that year, based on two-digit SIC industry) Q. Source: Compustat. ROA Return on assets is calculated as net income divided by total assets. Source: Compustat. Industry adj. ROA Firm ROA minus the median (from the Compustat population for that year, based on two-digit SIC industry) ROA. Source: Compustat. Interlock Firm A and B have interlocked directors if Firm A has Director 1 serving as an insider and Director 2 serving as an outsider, while Firm B has Director 1 serving as an outsider and Director 2 serving as an insider. Source: Corporate Library. Panel B. Board Structure and Composition Inside directors The sum of all directors currently employed by the company divided by total board size. Source: Corporate Library. Gray directors The sum of all directors who have either a current financial relationship with the company, or are former employees divided by total board size. Source: Corporate Library. CEO duality A binary variable equal to one if the CEO is also the current board chair and 0 otherwise. Source: Corporate Library. Board size The total number of directors on the board. Source: Corporate Library. Outside over 3 The proportion of outside directors that serve boards on four or more boards, relative to the total number of outside directors. Source: Corporate Library. Outside over 69 The proportion of all outside directors over age 69, relative to the total number of outside directors. Source: Corporate Library. Classified board A binary variable equal to one if the board is classified, and 0 otherwise. Source: Corporate Library. Panel C. CEO Characteristics and Ownership Structure CEO age The CEO's age as of the fiscal year end for each year. Source: Corporate Library. CEO tenure CEO tenure measures the number of years that the current CEO has been in that position. Source: Corporate Library. CEO ownership The proportion of shares held by the CEO divided by total shares outstanding. Source: Corporate Library. Inside ownership The percentage of equity held by top management and directors. Source: Corporate Library. Institutional The percentage of equity held by institutions. ownership Source: Corporate Library. Panel D. Other Control Variables Governance A binary variable equal to one if the firm has a policy explicit governance policy. Source: Corporate Library. Three-year sales The compounded annual growth rate of sales over CAGR the prior three years. Source: Corporate Library. Y02 A binary variable equal to one if the year equals 2002. Y03 A binary variable equal to one if the year equals 2003. Long-term debt Long-term debt divided by total assets. Source: Compustat. FIN A binary variable equal to one if the firm is in the financial sector (SIC code 6000-6999) and zero otherwise. Source: Compustat. Sales Sales revenue in $ millions. Source: Compustat.
Achen, C., 1986, The Statistical Analysis of Quasi-Experiments, Berkeley, CA, University of California Press.
Agrawal, A. and C. Knoeber, 1996, "Firm Performance and Mechanisms to Control Agency Problems between Managers and Shareholders," Journal of Financial and Quantitative Analysis 31, 377-397.
Amemiya, T., 1978, "The Estimation of a Simultaneous Equation Generalized Probit Model," Econometrica 46, 1,193-1,205.
Baliga, B., R. Moyer, and R. Rao, 1996, "CEO Duality and Firm Performance: What's the Fuss?" Strategic Management Journal 17, 41-43.
Barnea, A. and I. Guedj, 2006, "But Mom, All the Kids Have One! CEO Compensation and Director Networks," University of Texas Working Paper.
Bazerman, M. and F. Schoorman, 1983, "A Limited Rationality Model of Interlocking Directorates," Academy of Management and Review 8, 206-217.
Bebchuk, L. and A. Cohen, 2005, "The Costs of Entrenched Boards," Journal of Financial Economics 78, 409-433.
Boehmer, E., J. Musumeci, and A. Poulsen, 1991, "Event-Study Methodology under Conditions of Event-Induced Variance," Journal of Financial Economics 30, 253-272.
Bowen, R., S. Rajgopal, and M. Venkatachalam, 2008, "Accounting Discretion, Corporate Governance and Firm Performance," Contemporary Accounting Research 25, 351-405.
Byrd, J. and K. Hickman, 1992, "Do Outside Directors Monitor Managers? Evidence from Tender Offer Bids," Journal of Financial Economics 32, 195-222.
Coles, J., N. Daniel, and L. Naveen, 2006, "Managerial Incentives and Risk-Taking," Journal of Financial Economics 79, 431-468.
Core, J. and W. Guay, 1999, "The Use of Equity Grants to Manage Optimal Equity Incentive Levels," Journal of Accounting and Economics 28, 151-184.
Core, J. and W. Guay, 2002, "Estimating the Value of Employee Stock Option Portfolios and their Sensitivities to Price and Volatility," Journal of Accounting Research 40, 613-630.
Core, J., R. Holthausen, and D. Larcker, 1999, "Corporate Governance, Chief Executive Officer Compensation, and Firm Performance," Journal of Financial Economics 51, 371-406.
Corrado, C., 1989, "A Non-Parametric Test for Abnormal Security-Price Performance in Event Studies," Journal of Financial Economics 23, 385-395.
Cotter, J., A. Shivdasani, and M. Zenner, 1997, "Do Independent Directors Enhance Target Shareholder Wealth During Tender Offers?" Journal of Financial Economics 43, 195-218.
Coughlan, A. and R. Schmidt, 1985, "Executive Compensation, Management Turnover, and Firm Performance: An Empirical Investigation," Journal of Accounting and Economics 7, 43-66.
Davidson, R. and J. MacKinnon, 2004, Estimation and Inference in Econometrics, New York, Oxford University Press.
Erickson, M., M. Hanlon, and E. Maydew, 2006, "Is There a Link Between Executive Compensation and Accounting Fraud?" Journal of Accounting Research 44, 113-145.
Faleye, O., 2007, "Classified Boards, Firm Value and Managerial Entrenchment," Journal of Financial Economics 83, 501-529.
Fama, E. and K. French, 1992, "The Cross-Section of Expected Stock Returns," Journal of Finance 47, 427-465.
Fama, E. and M. Jensen, 1983, "Separation of Ownership and Control," Journal of Law and Economics 26, 301-326.
Ferris, S., M. Jagannathan, and A. Pritchard, 2003, "Too Busy to Mind the Business? Monitoring by Directors with Multiple Board Appointments," Journal of Finance 58, 1087-1112.
Fich, E. and A. Shivdasani, 2006, "Are Busy Boards Effective Monitors?" Journal of Finance 61, 689-724.
Fich, E. and L. White, 2005, "Why Do CEOs Sit on Each Other's Boards?" Journal of Corporate Finance 11, 175-195.
Greene, W., 2000, Econometric Analysis, Upper Saddle River, NJ, Prentice Hall.
Hallock, K., 1997, "Reciprocally Interlocking Boards of Directors and Executive Compensation," Journal of Financial and Quantitative Analysis 32, 331-344.
Huson, M., R. Parrino, and L. Starks, 2001, "Internal Monitoring Mechanisms and CEO Turnover: A Long-Term Perspective," Journal of Finance 56, 2265-2297.
Jensen, M., 1986, "Agency Costs of Free Cash Flow, Corporate Finance and Takeovers," American Economic Review 76, 323-339.
Jensen, M. and W. Meckling, 1976, "Theory of the Firm: Managerial Behavior, Agency Costs, and Ownership Structure," Journal of Financial Economics 3, 305-360.
Jensen, M. and K. Murphy, 1990, "Performance Pay and Top Management Incentives," Journal of Political Economy 98, 225-264.
John, K. and L. Senbet, 1998, "Corporate Governance and Board Effectiveness," Journal of Banking and Finance 22, 371-403.
Klein, A., 1998, "Firm Performance and Board Committee Structure," Journal of Law and Economics 41, 275-303.
Koenig, T. and R. Gogel, 1981, "Interlocking Corporate Directorships as a Social Network," American Journal of Economics and Sociology 40, 37-50.
Larcker, D., S. Richardson, A. Seary, and I. Tuna, 2006, "Director Networks, Executive Compensation and Firm Performance," University of Pennsylvania Working Paper.
Maddala, G., 1983, Limited-Dependent and Qualitative Variables in Econometrics, New York, Cambridge University Press.
McConnell, J. and H. Servaes, 1990, "Additional Evidence on Equity Ownership and Corporate Value," Journal of Financial Economics 27, 595-612.
Mizruchi, M. and L. Stearns, 1988, "A Longitudinal Study of the Formation of Interlocking Directorates," Administrative Science Quarterly 33, 194-210.
Modigliani, E and M. Miller, 1963, "Corporate Income Taxes and the Cost of Capital: A Correction," American Economic Review 53, 433-443.
Nguyen-Dang, B., 2007, "Does the Rolodex Matter? Corporate Elite's Small World and Effectiveness of Boards of Directors," Chinese University of Hong-Kong Working Paper.
Pfeffer, J. and G. Salancik, 1978, The External Control of Organizations, New York, Harper & Row.
Raghavan, A., 2005, "More CEOs Say 'No Thanks' to Board Seats," Wall Street Journal January 28, B1.
Richardson, J., 1987, "Directorship Interlocks and Corporate Profitability," Administrative Science Quarterly 32, 367-386.
Rosenstein, S. and J. Wyatt, 1990, "Outside Directors, Board Independence, and Shareholder Wealth," Journal of Financial Economics 26, 175-191.
Shivdasani, A. and D. Yermack, 1999, "CEO Involvement in the Selection of New Board Members: An Empirical Analysis," Journal of Finance 54, 1829-1853.
Tong, Z., 2008, "Deviation from Optimal CEO Ownership and Firm Value," Journal of Banking and Finance 32, 2462-2470.
Useem, M., 1982, "Classwide Rationality in the Politics of Managers and Directors of Large Corporations in the United States and Great Britain," Academy of Management Review 6, 243-252.
Vafeas, N., 1999, "Board Meeting Frequency and Firm Performance," Journal of Financial Economics 53, 113-142.
Yermack, D., 1996, "Higher Market Valuation of Companies with a Small Board of Directors," Journal of Financial Economics 40, 185-211.
The authors thank participants at the 2005 Financial Management Association annual meeting, especially the discussant Ken Ayotte, and an anonymous referee for invaluable comments and suggestions. The authors gratefully acknowledge financial assistance from College of Business (Ohio University) research seed grant program and the Gardner Fellowship.
Erik Devos, Andrew Prevost, and John Puthenpurackal*
* Erik Devos is an Associate Professor of Finance at the University of Texas in El Paso, TX. Andrew Prevost is an Associate Professor of Finance at Ohio University in Athens, OH. John Puthenpurackal is an Associate Professor of Finance at the University of Nevada in Las Vegas, NV.
(1) See US Proxy Manual (Institutional Shareholder Services, 2004) for details on Securities and Exchange Commission (SEC) disclosure requirements when interlocked directorships exist.
(2) The Corporate Library collects and maintains a large data set containing information on an array of governance and other variables for a large sample of firms. See www.thecorporatelibrary.com.
(3) For example, it is not clear that two directors who serve together on boards of two firms as outsiders create anyconflict of interest that would result in reduced monitoring effectiveness.
(4) See Davidson and MacKinnon (2004) for specific details.
(5) The results are qualitatively similar when matching on prior performance and industry, and prior performance and size.
(6) Larcker et al. (2006) and Barnea and Guedj (2006) use different measures of interrelatedness and finds that CEOs of firms whose directors are more connected tend to have excess pay.
(7) As discussed by Core and Guay (1999, p. 155), other researchers (Jensen and Murphy, 1990) focus on changes in wealth for a dollar (vs. percentage) change in firm value. The delta measure is equivalent to the Jensen and Murphy (1990) measure multiplied by the market value of the firm and divided by $100,000.
(8) See Core and Guay (1999) for details of the variables used and the hypotheses regarding these variables.
(9) Merging the sample used in the prior analysis with Execucomp results in the loss of 624 observations (see Table I).
(10) See Huson, Parrino, and Starks (2001) for further details of this turnover classification scheme.
Table I. Sample Selection Criteria Firm-Year Selection Criteria Observations Firms in Corporate Library (2001-2003) 5,302 Less: Nonordinary shares (10) Less: Foreign firms (259) Less: Subsidiaries (52) Eligible Corporate Library data 4,981 Less: Cusip not on Compustat (688) Less: Outsider interlocks (330) Less: Missing Compustat control variables (397) Final sample (performance analysis) 3,566 Less: Missing Execucomp CEO compensation data (624) Final sample (excess pay-performance sensitivity) 2,942 Table II. Interlocked Director Characteristics Panel A reports the number of interlocks in each year of our sample period, Panel B gives the time trend of when interlocks appear for the first time in the sample period, and Panel C provides additional details about the interlocked firms and the directors comprising the interlock. Proportion Number of Sample Panel A. Time Trend of Director Interlocks In 2001 71 0.511 In 2002 39 0.281 In 2003 29 0.209 Total 139 1.000 Panel B. Time Trend of New Director Interlocks First year is 2001 71 0.755 First year is 2002 11 0.117 First year is 2003 12 0.128 Total 94 1.000 Panel C. Interlock Characteristics (out of 94) Same industry (1-digit SIC) 11 0.117 Financial firm 30 0.319 Interlocked with 31 0.330 a financial firm Insider is CEO of firm 32 0.340 Interlocked with CEO 37 0.394 Table III. Summary Statistics of Governance and Financial Characteristics This table presents descriptive statistics of variables used in the performance analysis for 3,566 sample firm-years over the 2001-2003 time period. Variable definitions are provided in the Appendix. Mean Std. Dev. Q1 Panel A. Endogenous Variables Interlock 0.033 0.179 0.000 Q 1.933 1.531 1.099 Industry adj. Q 0.545 1.458 -0.142 ROA 0.017 0.199 0.008 Industry adj. ROA 0.040 0.207 0.002 Panel B. Board Structure and Composition Inside directors 0.228 0.136 0.125 Gray directors 0.083 0.123 0.000 CEO duality 0.697 0.460 0.000 Board size 9.346 3.042 7.000 Outside over 3 boards 0.138 0.165 0.000 Outside over 69 0.106 0.154 0.000 Classified board 0.607 0.489 0.000 Panel C. CEO Characteristics and Ownership Structure CEO age 54.765 7.380 50.000 CEO tenure 7.811 8.055 2.000 CEO ownership 0.027 0.068 0.001 Inside ownership 0.220 0.185 0.060 Institutional ownership 0.619 0.201 0.486 Panel D. Other Control Variables Governance policy dummy 0.125 0.331 0.000 Three-year sales CAGR 0.107 0.195 0.005 Long-term debt 0.188 0.183 0.027 FIN 0.143 0.350 0.000 Sales ($MM) 4,507.161 13,325.291 464.030 Median Q3 Panel A. Endogenous Variables Interlock 0.000 0.000 Q 1.403 2.110 Industry adj. Q 0.100 0.700 ROA 0.035 0.075 Industry adj. ROA 0.034 0.104 Panel B. Board Structure and Composition Inside directors 0.182 0.286 Gray directors 0.000 0.143 CEO duality 1.000 1.000 Board size 9.000 11.000 Outside over 3 boards 0.100 0.250 Outside over 69 0.000 0.175 Classified board 1.000 1.000 Panel C. CEO Characteristics and Ownership Structure CEO age 55.000 59.000 CEO tenure 5.000 11.000 CEO ownership 0.003 0.016 Inside ownership 0.180 0.341 Institutional ownership 0.638 0.776 Panel D. Other Control Variables Governance policy dummy 0.000 0.000 Three-year sales CAGR 0.078 0.178 Long-term debt 0.157 0.292 FIN 0.000 0.000 Sales ($MM) 1,109.800 3,237.200 Table IV. Univariate Comparisons of Characteristics of Interlocked and Noninterlocked Firm-Years This table shows mean and medians (in parentheses) for selected variables. Variable definitions are provided in the Appendix. Interlock Noninterlock (N=118) (Full Sample, N = 3,448) Panel A. Performance Variables Q 1.700 1.941 (1.213) (1.412) Industry- 0.409 0.549 adjusted Q (0.022) (0.103) ROA 0.020 0.016 (0.023) (0.035) Industry- 0.021 0.041 adjusted (0.014) (0.035) ROA Panel B. Board Structure and Composition Inside directors 0.202 0.229 (0.182) (0.182) Gray directors 0.041 0.084 (0.000) (0.000) CEO duality 0.695 0.771 (1.000) (1.000) Board size 11.695 9.266 (11.000) (9.000) Outside 0.237 0.134 over 3 boards (0.200) (0.100) Outside over 69 0.071 0.106 (0.000) (0.000) Classified board 0.723 0.602 (1.000) (1.000) Panel C. CEO Characteristics and Ownership Structure CEO age 56.254 54.685 (58.000) (55.000) CEO tenure 7.873 7.809 (5.000) (5.000) CEO ownership 0.023 0.027 (0.002) (0.003) Inside ownership 0.205 0.220 (0.150) (0.180) Institutional 0.548 0.622 ownership (0.545) (0.643) Panel D. Other Control Variables Governance 0.195 0.123 policy (0.000) (0.000) Three-year sales 0.077 0.109 CAGR (0.068) (0.079) Long-term debt 0.222 0.187 (0.203) (0.155) FIN 0.246 0.140 (0.000) (0.000) Sales ($MM) 11,568.710 4,265.496 (3,804,630) (1,086.819) Interlock- Noninterlock Noninterlock (Size- t- (z-) Statistic Constrained, N = 1,724) Panel A. Performance Variables Q -1.69 * 1.812 (-2.70) *** (1.369 Industry- -1.15 0.450 adjusted Q (-1.37) (0.096) ROA 0.21 0.034 (-1.15) (0.037) Industry- -1.02 0.048 adjusted (-2.67) *** (0.032) ROA Panel B. Board Structure and Composition Inside directors -2.06 *** 0.203 (-1.75) * (0.167) Gray directors -5.72 *** 0.081 (-3.71) *** (0.000) CEO duality -1.78 * 0.746 (-1.78) * (1.000) Board size 7.76 *** 10.175 (8.30) *** (10.000) Outside 5.64 *** 0.175 over 3 boards (6.44) *** (0.143) Outside over 69 -3.08 *** 0.090 (-2.13) ** (0.000) Classified board 2.77 *** 0.626 (2.76) *** (1.000) Panel C. CEO Characteristics and Ownership Structure CEO age 2.27 ** 55.372 (3.28) *** (55.000) CEO tenure 0.09 7.191 (0.57) (5.000) CEO ownership -0.74 0.021 (-1.36) (0.002) Inside ownership -0.87 0.198 (-1.45) (0.150) Institutional -4.70 *** 0.644 ownership (-4.51) *** (0.661) Panel D. Other Control Variables Governance 1.95 * 0.198 policy (2.33) ** (0.000) Three-year sales -2.65 *** 0.102 CAGR (-0.87) (0.073) Long-term debt 2.38 ** 0.214 (3.22) *** (0.197) FIN 2.63 *** 0.125 (3.23) *** (0.000) Sales ($MM) 3.39 *** 8,046.315 (7.53) *** (3,031.737) Interlock- Noninterlock t- (z-) Statistic Panel A. Performance Variables Q -0.90 (-2.28) ** Industry- -0.73 adjusted Q (-1.08) ROA -0.70 (-1.84) * Industry- -1.47 adjusted (-2.38) ** ROA Panel B. Board Structure and Composition Inside directors -0.02 (0.57) Gray directors -5.18 *** (-3.65)*** CEO duality -0.61 (-0.61) Board size 4.82 *** (4.95) *** Outside 3.34 *** over 3 boards (3.64) *** Outside over 69 -1.66 * (-1.50) Classified board 2.23 ** (2.23) ** Panel C. CEO Characteristics and Ownership Structure CEO age 1.32 (2.37) ** CEO tenure 0.91 (0.39) CEO ownership 0.41 (0.92) Inside ownership -0.42 (-0.12) Institutional -5.57 *** ownership (-5.72) *** Panel D. Other Control Variables Governance -0.09 policy (-0.09) Three-year sales -1.47 CAGR (-0.42) Long-term debt 0.51 (0.90) FIN 2.98 *** (3.75) *** Sales ($MM) 1.61 (0.05) *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level. Table V. Interlock-Firm Performance Analysis Using Two-Stage Probit Least Squares Estimation Variable definitions are provided in the Appendix. The t- (z)-statistics are based on cluster-adjusted robust standard errors. Model 1 Industry Adj. Q Coef. t-stat Intercept 0.252 0.26 Endogenous Variables Industry adj. Q Industry adj. ROA Interlock -0.233 -1.24 Board Structure and Composition Inside directors 0.811 ** 2.38 Gray directors -0.061 -0.21 CEO duality -0.002 -0.03 Log (Board size) -0.133 -0.59 Outside over 3 boards 0.850 *** 3.38 Outside over 69 -0.064 -1.19 Classified board -0.085 -1.19 CEO Characteristics and Ownership Structure CEO age CEO tenure CEO ownership Inside ownership 0.475 1.04 Inside ownership -0.624 -0.96 Institutional [ownership.sup.2] Other Control Variables Governance policy 0.141 * 1.94 Three-year sales CAGR 0.019 *** 6.59 Long-term debt -0.943 *** -3.93 Log (Sales) -0.015 -0.43 FIN Y02 -0.138 * -1.96 Y03 -0.261 *** -2.84 F-statistic (LR [chi square]) 12.20 Prob > F (Prob > [chi square]) 0.000 Adj [R.sup.2] (Pseudo [R.sup.2]) 0.146 No. obs. 3,566 No. clusters 1,463 Model 1 Interlock Coef. z-stat Intercept -3.569 *** -4.27 Endogenous Variables Industry adj. Q -0.359 *** -3.34 Industry adj. ROA Interlock Board Structure and Composition Inside directors Gray directors CEO duality Log (Board size) 0.629 *** 2.70 Outside over 3 boards 1.260 *** 3.89 Outside over 69 -0.598 -1.41 Classified board CEO Characteristics and Ownership Structure CEO age -0.000 -0.01 CEO tenure 0.013 1.37 CEO ownership 0.369 0.46 Inside ownership Inside ownership Institutional [ownership.sup.2] -0.718 ** -2.29 Other Control Variables Governance policy 0.028 0.19 Three-year sales CAGR Long-term debt Log (Sales) 0.108 ** 2.52 FIN 0.097 0.75 Y02 -0.361 *** -4.23 Y03 -0.454 *** -4.09 F-statistic (LR [chi square]) (92.04) Prob > F (Prob > [chi square]) (0.000) Adj [R.sup.2] (Pseudo [R.sup.2]) (0.146) No. obs. 3,566 No. clusters 1,463 Model 2 Industry Adj. Q Coef. t-stat Intercept -0.579 *** -4.3 Endogenous Variables Industry adj. Q Industry adj. ROA Interlock -0.088 *** -4.37 Board Structure and Composition Inside directors 0.017 0.59 Gray directors -0.179 *** -4.04 CEO duality 0.005 0.51 Log (Board size) 0.079 *** 2.64 Outside over 3 boards 0.046 1.35 Outside over 69 -0.02 -0.67 Classified board 0.020** 2.18 CEO Characteristics and Ownership Structure CEO age CEO tenure CEO ownership Inside ownership 0.189 ** 2.55 Inside ownership -0.157 * -1.78 Institutional [ownership.sup.2] Other Control Variables Governance policy -0.021 ** -1.99 Three-year sales CAGR 0.002 *** 4.56 Long-term debt -0.113 *** -3.01 Log (Sales) 0.032 *** 4.63 FIN Y02 -0.009 -0.95 Y03 -0.000 -0.00 F-statistic (LR [chi square]) 11.430 Prob > F (Prob > [chi square]) 0.000 Adj [R.sup.2] (Pseudo [R.sup.2]) 0.100 No. obs. 3,566 No. clusters 1,463 Model 2 Interlock Coef. z-stat Intercept -4.854 *** -5.75 Endogenous Variables Industry adj. Q Industry adj. ROA -2.225 ** -2.44 Interlock Board Structure and Composition Inside directors Gray directors CEO duality Log (Board size) 0.791 *** 3.50 Outside over 3 boards 0.981 *** 3.00 Outside over 69 -0.608 -1.45 Classified board CEO Characteristics and Ownership Structure CEO age 0.008 0.80 CEO tenure 0.007 0.81 CEO ownership 0.508 0.66 Inside ownership Inside ownership Institutional [ownership.sup.2] -0.588 * -1.86 Other Control Variables Governance policy -0.061 -0.42 Three-year sales CAGR Long-term debt Log (Sales) 0.158 *** 3.37 FIN 0.036 0.25 Y02 -0.299 *** -3.75 Y03 -0.368 *** -3.54 F-statistic (LR [chi square]) (93.00) Prob > F (Prob > [chi square]) (0.000) Adj [R.sup.2] (Pseudo [R.sup.2]) (0.142) No. obs. 3,566 No. clusters 1,463 *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level. Table VI. Univariate Comparison of Performance Metrics of Interlocked Sample and Control Samples This table presents univariate performance comparisons of sample firms relative to control firms matched on prior performance (one-year lagged ROA), size (total assets), and industry (two-digit SIC). Panel A reports performance differences using the sample time period (2001-2003), and Panel B reports performance differences using the interlock formation year. Panel A. Matched Sample Based on Prior Performance, Industry, and Size (2001-2003 Sample Period) Interlock Noninterlock Interlock- (N = 118) (N = 118) Noninterlock Q 1.700 1.747 -0.047 (1.213) (1.251) (-0.015) Industry-adjusted Q 0.409 0.406 0.003 (0.022) (0.056) (0.000) ROA 0.020 0.027 -0.007 (0.023) (0.033) (-0.003) * Industry-adjusted ROA 0.021 0.036 -0.015 * (0.014) (0.020) (-0.004) * Panel B. Matched Sample Based on Prior Performance, Industry, and Size (Formation Year of Interlock) Interlock- Interlock Non Interlock Non Interlock (N = 70) (N = 70) Q 2.000 1.644 0.356 (1.348) (1.255) (-0.013) Industry-adjusted Q 0.661 0.257 0.403 (0.105) (0.057) (0.023) ROA 0.049 0.033 0.016 ** (0.039) (0.029) (0.003) * Industry-adjusted ROA 0.035 0.021 0.014 (0.012) (0.057) (0.002) ** Significant at the 0.05 level. * Significant at the 0.10 level. Table VII. Stock Price Reaction to the Creation of Interlocked Directors This table reports (-1, +1) and (-1, +10) cumulative abnormal returns (CARs) surrounding the announcement dates that create interlocked directors, for various samples. Equally weighted (first column) and median (second column) event-period CARs are measured using the CRSP equal-weighted index with Scholes-Williams betas. The market model parameters are estimated over the preevent estimation period extending from day -221 to day -21. The number of positive and negative individual excess returns is given in the third column. Parametric z-statistics and nonparametric rank test statistics for the significance of the CARs are provided in the fourth and fifth columns, respectively. Window Mean CAR Median CAR Pos: Neg StdCsect z Rank Test z Panel A. Full Sample (N= 69) (-1, +1) -0.35 -0.12 34 : 35 -0.81 -0.62 (-1, +10) -1.67 -1.26 22 : 47 -2.25 ** -1.97 ** Panel B. Directors i and j Already on Board A (n = 28) (-1, +1) -1.58 -1.46 12 : 16 -2.59 *** -2.58 ** (-1, +10) -1.40 -1.49 10 : 18 -1.24 -1.79 * Panel C. Director i or j New to Board A (n = 41) (-1, +1) 0.48 0.16 22 : 19 0.79 0.91 (-1, +10) -1.85 -1.23 12 : 29 -1.88 * -1.09 Panel D. Director i or j Is CEO on Board A (n = 32) (-1, +1) -0.61 -0.33 15 : 17 -0.67 -0.27 (-1, +10) -1.95 -1.14 11 : 21 -2.27 ** -1.11 Panel E. Directors i and j Are Non-CEO on Board A (n = 37) (-1, +1) -0.13 0.05 19 : 18 -0.50 0.44 (-1, +10) -1.42 -1.81 11 : 26 -1.29 -2.19 ** *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level. Table VIII. Interlocked Directors and Optimal Equity Ownership Incentives Panel A provides univariate comparisons for CEO and management team equity incentive portfolio delta and residual delta for interlocked and noninterlocked firms using the difference-in-means (Wilcoxon) test for means (medians). Panel B reports least squares estimates using residual portfolio equity incentives for the CEO (Column 1) and management team (Column 2), respectively, as the dependent variable. White heteroskedastic-consistent cluster-adjusted p-values are reported in parentheses. Variable definitions are provided in the Appendix. Panel A. Univariate Comparisons of Delta and Residual Delta Interlocked No. Obs. Non- interlocked CEO portfolio delta 3,298.342 113 3,723.035 (755.002) (540.480) Management team 7,582.619 114 5,983.765 portfolio delta (1,893.904) (1,251.005) CEO residual -2,523.09 110 92.023 portfolio delta (-3,211.96) (-1,277.040) Management team -2,499.44 111 91.350 residual portfolio delta (-4,332.95) (-1,039.780) Panel A. Univariate Comparisons of Delta and Residual Delta No. Obs. P-value for Difference in Mean (Median) CEO portfolio delta 3,181 0.670 (0.000) Management team 3,190 0.549 portfolio delta (0.000) CEO residual 3,016 0.018 portfolio delta (0.031) Management team 3,026 0.349 residual portfolio delta (0.023) Panel B. Regressions of Residual Delta on Interlock, Board, and Ownership Structure Characteristics CEO Residual Management Team Residual Portfolio Delta Portfolio Delta Intercept 12,821.308 31,363.320 (2.48) (2.37) Interlock -5,503.163 -6,298.817 (-2.32) (-1.53) CEO duality -874.043 -3,268.884 (-0.97) (-1.54) Board size -498.670 -1,290.556 (-2.70) (3.04) Inside directors 13,494.677 13,094.000 (1.00) (0.96) Gray directors -4,891.694 -4,251.402 (-1.42) (-0.97) Outside over 3 boards 20,234.884 35,056.000 (1.67) (1.93) Outside over 69 -662.836 -2,269.576 (-0.24) (-0.51) Insider ownership 241.896 270.420 (2.22) (2.06) Institutional ownership -209.066 -378.498 (-3.74) (-3.05) F-value 7.97 12.01 (p-value) (0.000) (0.000) Adj. [R.sup.2] 0.021 0.033 No. obs. 2,931 2,942 No. clusters 1,264 1,268 Table IX. Interlocked Directors and CEO Turnover This table presents a series of probit regressions with HPS CEO turnover and All CEO turnover (measures of CEO turnover) as dependent variables. HPS CEO turnover equals one if the CEO departure is forced based on the Huson, Parrino, and Starks (2001) algorithm and zero otherwise. All CEO turnover equals one if the CEO changes for any reason and zero otherwise. Market adj. [return.sub.t-1] is defined as the firm's stock return over the prior year minus the S&P 500 return over that period. Proportion outside directors is the number of outside directors divided by total board size. Other variable definitions are provided in the Appendix. White heteroskedastic- consistent cluster-adjusted p-values are reported in parentheses. Independent Variable HPS CEO Turnover All CEO Turnover Intercept -8.895 -8.163 (-5.10) (-9.10) Market adj. [return.sub.t-1] -0.026 -0.008 (-4.05) (-4.28) Interlock 0.001 0.089 (0.00) (0.765) Interlock x Market adj. 0.045 0.017 [return.sub.t-1] (0.64) (2.48) Proportion outside directors 0.593 0.572 (0.67) (1.33) CEO duality -0.316 -1.777 (-3.96) (-12.64) Log (Board size) 0.086 0.207 (0.11) (0.78) Outside over 3 boards -0.273 0.999 (-0.33) (2.36) Outside over 69 -1.748 -1.229 (-1.45) (-2.73) CEO age 0.024 0.107 (1.41) (10.24) Institutional ownership 1.320 0.403 (1.38) (1.15) Log (Sales) 0.302 0.050 (2.80) (0.85) Y02 0.681 -1.437 (1.59) (-7.46) Y03 0.586 0.050 (1.58) (0.37) Wald [chi square] 62.86 259.52 (Prob > [chi square]) (0.000) (0.000) Pseudo [R.sup.2] 0.132 0.206 No. obs. 2,819 2,819 No. clusters 1,185 1,185 No. turnovers 49 344
|Printer friendly Cite/link Email Feedback|
|Author:||Devos, Erik; Prevost, Andrew; Puthenpurackal, John|
|Article Type:||Statistical table|
|Date:||Dec 22, 2009|
|Previous Article:||Investment timing for dynamic business expansion.|
|Next Article:||An empirical comparison of option-pricing models in hedging exotic options.|