Is There Performance-Based Turnover on Corporate Boards?

Transcription

1 Is There Performance-Based Turnover on Corporate Boards? Thomas W. Bates David A. Becher Jared I. Wilson October 2015 Abstract We examine the threat of turnover as an incentive to align corporate directors interests with those of their shareholders. Our results suggest an economically significant relation between director turnover and prior firm performance that only manifests in idiosyncratic stock returns; consistent with the monitoring of actions attributable to management. Director turnover-performance sensitivity is higher post-sox and is moderated by governance structures. The incentive effect of turnover is partially offset by the inability of poorly performing firms to attract higher value human capital. Overall, the threat of director turnover related to firm performance has meaningful implications for the director labor market. Keywords: Director turnover, firm performance, board seats, replacement directors JEL Classification: G34, J23 Thomas Bates: Arizona State University, W.P. Carey School of Business, phone: (480) , David Becher: Drexel University, LeBow College of Business, and Fellow, Wharton Financial Institutions Center, University of Pennsylvania, phone: (215) , Jared Wilson: Drexel University, LeBow College of Business, We thank Laura Lindsey, Michelle Lowry, Ralph Walkling as well as seminar participants at Drexel University, the University of Delaware, and the University of Mississippi for their helpful comments. 1

2 1. Introduction The fiduciary role of the corporate board of directors is to monitor and advise management and, more generally, to represent the interests of shareholders in a firm s business dealings (e.g. Fama, 1980, Fama and Jensen, 1983). Over the past three decades, boards of public corporations have faced increasing scrutiny over their effectiveness by regulators, institutional investors, and other stakeholders (e.g., Parrino, Sias, and Starks, 2003; Linck, Netter, and Yang, 2009). This heightened scrutiny is largely predicated on the expectation that directors have a significant influence over firm decision making and performance. We consider whether the directors of public corporations systematically face the threat of turnover from their board positions for past poor performance. Agency theory posits that directors should be exposed to the threat of replacement for poor performance as an incentive mechanism to align their interests with those of shareholders. Empirically, a number of studies find a negative turnover-performance relation for executives of public firms, however, there is no evidence that board members similarly experience turnover after poor performance. It is possible that there is no relation between director turnover and past performance given that board members themselves exercise a substantial degree of discretion in determining board composition and their own tenure. In keeping with this notion, Yermack (2004) suggests that no higher authority exists to discipline established directors for poor performance. In addition, Fahlenbrach, Low, and Stulz, (2014) indicate that reputational costs may motivate some directors to resign prior to the realization of poor firm performance, mitigating the sensitivity of turnover to past firm performance. In this study, we examine the incidence of turnover for a broad sample of non-executive directors in U.S. public firms between 2000 and 2011; a period that saw dramatic changes in the composition of, and scrutiny over, corporate boards. Our results indicate that directors are more 2

3 likely to turnover after the realization of poor stock and accounting performance. The relation between firm performance and director turnover obtains both for individual directors and at the firm level. For example, directors of firms in the lowest quartile of industry-adjusted stock returns in a given year are 1.1% more likely to turnover in the next year. Similarly, for directors of firms in the lowest quartile of industry-adjusted accounting returns, the likelihood of turnover the next year is 1.8% higher. These effects are economically significant given an unconditional annual turnover rate of 7.6% for the directors in our sample. Following Bertrand and Mullainathan (2001), we decompose stock returns into luck and skill and find that the turnover-performance relation manifests only in the idiosyncratic component of returns (skill). This suggests that director turnover is a consequence of poor performance directly attributable to a firm s management, and by extension the board, rather than a firm s industry or overall market performance. To provide further evidence on the disciplinary nature of director turnover, we investigate the ex-post outcomes for directors that turnover. For the threat of replacement to be a meaningful incentive, we should observe ex-post settling up of the director labor market. To test this, we examine the change in the number of board seats held by directors as well as the probability that a director obtains an additional board seat over the three-year window after turnover is identified. Directors are more than 40% less likely to gain a new directorship following their turnover, compared to directors that do not turnover; a relationship that only manifests for directors that turnover in the context of poor firm performance. These findings suggest that director turnover is in fact disciplinary as these directors are consequently penalized in the labor market. One alternative explanation for the observed negative relation between director turnover and firm performance is that directors may voluntarily step down to preserve their remaining reputational capital. Given the structure of our tests, it is likely too late for directors to detach 3

4 themselves from already observed poor firm performance. However, some directors may be able to anticipate further poor firm performance and exit to avoid additional reputational damage. To consider this possibility we examine the performance of firms after director departures. Contrary to the reputation conjecture above, we find that firm performance substantially improves after directors depart the firm following poor performance. Another alternative explanation for our findings is that highly qualified directors may voluntarily exit to provide a costly, yet credible, signal of quality by disassociating themselves from the poorly performing firm. If this is the case, we expect that these directors should be more likely to experience a net gain in future directorships relative to the directors of the firm that do not to leave the board. Contrary to the signaling conjecture, however, we find that departing directors actually experience a net loss in the number of other directorships they hold in the future. Finally, we consider if our findings are a mainly byproduct of coincidental turnover between CEOs and directors as documented in Farrell and Whidbee (2000). This is not the case, however, as we document an economically significant relation between turnover and firm performance for outside directors even in the absence of coincident CEO turnover. Linck, Netter and Yang (2009) assert that scrutiny over directors actions increased following the implementation of SOX. Examining changes in the director turnover-performance sensitivity over the time series is important given the structural shift towards corporate boards comprised largely of unaffiliated outside directors and the contemporaneous increase in director accountability. 1 We document that the negative relation between director turnover and stock price 1 For example, anecdotal evidence suggests that activist funds have been increasingly involved in proxy contests to remove directors of firms they claim are underperforming. Public contests are costly and infrequent, and have been met with mixed success. For example, in a 2014 proxy fight, Starboard replaced the board of Darden Restaurants with a slate of directors backed by its fund. In 2015, Trian Fund Management failed to replace directors of DuPont with its own slate. See J. Jargon, J. Lublin, and D. Benoit, Starboard Succeeds in Replacing Entire Darden Board The Wall Street Journal, October 10, 2014; and J. Bunge and D. Benoit, DuPont Defeats Peltz in Board Fight The Wall Street Journal, May 14,

5 performance, specifically the skill component of returns, more than twice as strong following the implementation of SOX and contemporaneous exchange listing requirements suggesting that the the director labor market is more dynamic, and increasingly tied to firm performance, in the post- SOX era. Prior research including Weisbach (1988) and Denis, Denis and Sarin (1997) indicates that corporate governance can moderate the turnover-performance sensitivity for executives, however, the role of governance in turnover for non-executive directors is unknown. As noted in Yermack (2004), it is unclear who, if anyone, monitors these monitors. We find that governance characteristics have a direct impact on director turnover. Directors are less likely to turnover in firms that exhibit weak internal governance such CEO-chair duality or when the directors have joined the board during the tenure of the sitting CEO and are thus perceived as being co-opted. Conversely, directors have a higher likelihood of turnover at firms with an active outside blockholder. The director turnover-performance sensitivity is also moderated by select governance characteristics. For example, directors on busy boards are 1.6% less likely to turnover after poor stock performance relative to directors not on busy boards, while the presence of an active blockholder increases the likelihood of director turnover by 1.3% in the year following relatively poor stock performance. Overall, these results suggest governance plays an important role in insulating or exposing directors to the threat of turnover. Given a significant relation between director turnover and firm performance, we consider the economic consequences for firms experiencing departures by examining the quality of directors appointed after turnover. Overall, replacement directors are more experienced than the directors that tunover; on average, they hold more directorships and are more likely to be a CEO elsewhere compared to directors that left the board. Given the importance of reputation in the 5

6 director labor market, we posit that firms with poor performance and coincident director turnover will find it difficult to hire qualified directors in the near future. Consistent with this, higher performing firms are able to attract better quality replacements relative to the directors that left the firm, while firms with poor prior performance, on average, hire directors that do not possess superior characteristics relative to the directors that they replace. These results suggest that poorly performing firms are handicapped in the director labor market, perversely just at the time when they would most benefit from director experience and skill. Our evidence concerning director turnover-performance sensitivity builds on prior studies that examine performance incentives for outside directors and managers. Fama (1980) and Fama and Jensen (1983) argue that outside directors have an incentive to be effective monitors to signal their value to shareholders and the outside labor market. The results from this study suggest that the threat of turnover for poor firm performance is a significant incentive for individual directors with meaningful economic consequences in the director labor market. 2. Related Literature Our paper builds on the prior literature that examines how incentive mechanisms in the labor market can align the interests of agents, including corporate directors, with those of shareholders. Yermack (2004) notes that turnover and direct compensation, such as stock and option grants, provide the most significant economic incentives for directors. He considers the relation between director turnover and firm performance for 734 Fortune 500 directors with tenure of five or less years and finds a negative relation between director turnover and contemporaneous stock performance, but no relation for accounting or lagged performance. Notably, this turnoverperformance sensitivity obtains only for directors in their first two years of service on the board, but not for directors with longer tenures. Given his mixed evidence, Yermack (2004) concludes 6

7 that for outside directors, the threat of replacement is more attenuated, since directors do not report to a higher authority that might fire them for poor performance. Fahlenbrach et al. (2014) argue that it is the directors, themselves, who decide when to continue and when to resign. Their empirical results suggest that some directors anticipate future poor performance and step down in advance to protect their reputation. This study examines the relationship between performance and the turnover of outside and unaffiliated directors between 2000 and 2011; a period that coincides with the enactment of SOX. Linck et al. (2009) note that SOX increased the workload for directors and heightened public scrutiny over their decisions. In addition, Guo and Masulis (2015) and Linck et al. (2009) document that SOX and changes in listing requirements necessitated a substantial rise in the proportion of outside directors on the boards of public corporations. Our work is closely related to the literature concerning turnover and incentives for corporate executives. For example, Weisbach (1988) and Parrino (1997) find a negative relation between firm performance and CEO turnover. Kaplan and Minton (2012) and Peters and Wagner (2014) document that CEO turnover and the sensitivity of executive turnover to firm performance has increased since Other authors focus on director departures in a variety of different contexts. For example, Harford (2003) finds that a majority of target directors lose their board seat after completed deals. Farrell and Whidbee (2000) show an increased likelihood of outside director turnover following forced CEO turnover. Fos and Tsoutsoura (2014) provide evidence that directors are likely to lose their board seat following a proxy contest, and Ertimur, Ferri, and Maber (2012) note that compensation committee members of firms that engaged in option backdating experience more turnover than non-compensation committee members. 7

8 An additional contribution of this paper is the examination of consequences of turnover for departing directors and the firms they leave. To the best of our knowledge, we are the first to examine various aspects of the quality of directors appointed to firms after director turnover. In addition, we focus on ex-post settling-up in the labor market post-turnover. If the labor market plays a role in incentivizing directors to act in shareholders best interest, it is important to understand the magnitude of this incentive. In related work, Yermack (2004) documents a positive relation between performance and the number of board seats obtained by outside directors in the future. Coles and Hoi (2003) find that directors who rejected provisions of Pennsylvania Senate Bill 1310 were more likely to gain additional directorships. Fich and Shivdasani (2007) show that directors of firms facing lawsuits experience a decline in the number of other board seats they hold. Ertimur, Ferri, and Maber (2015) document that directors decrease their number of additional directorships after proxy contests. Finally, the prior literature establishes a positive relation between the experience of an individual and the likelihood that they are appointed to a board. For example, Fahlenbrach, Low and Stulz (2010) find that CEO experience is an important determinant of director appointments. In addition, Harford and Schonlau (2013) find that the likelihood of future directorships is positively correlated with experience for directors in acquiring firms. 3. Data and summary statistics The initial sample of director data is drawn from Management Diagnostic s BoardEx database, which includes 430,993 director-firm-year observations during our sample period that starts at the beginning of 2000 and ends in We merge our sample of director-firm-year observations with Compustat to obtain firm-level accounting data, with the Center for Research of Stock Prices (CRSP) database for firm stock returns, and with institutional ownership data from the Thomson Reuters Institutional Ownership database. After excluding director-firm-year 8

9 observations with missing values for returns, book value of assets and institutional ownership, the sample consists of 388,461 director-firm-year observations. To identify director turnover events, we follow a director from one firm-year board report date on BoardEx to the next, where a report date corresponds to a firm s fiscal year end. Directors that are no longer listed at a subsequent report date are classified as turnover directors, while those who remain listed are classified as non-turnover directors. We exclude turnover attributable to a director s death (reported by BoardEx). We also eliminate 40,465 firm-year observations where there is no subsequent board report to identify turnover. As we require a follow-on report date, none of our turnover observations are due to acquisition, delisting, or privatization. To focus on a sample of outside director turnover, we remove 64,723 director-firm-year observations where the director is also an officer of the firm. BoardEx provides information on a variety of director characteristics including age, tenure, committee membership, and past and current employment and directorships. We delete 2,211 director-firm-year observations with missing values for age and tenure as these are essential control variables for this study. The final sample consists of 281,062 director-firm-year observations and 21,275 observations of director turnover yielding an unconditional turnover rate of 7.57%. The sample includes 5,802 unique firms and 43,351 distinct directors. At the firm level, the sample has 39,975 firm-year observations where 14,220 are associated with at least one director turnover. Panel A of Table 1 summarizes four measures of one-year lagged firm performance. 2 Industry-adjusted stock return is a sample firm s annual buy-and-hold return minus the annual buy-and-hold return for the median firm in the same Fama-French 48 industry. Industry-adjusted ROA is calculated as a sample firm s net income scaled by total book value of assets, minus the 2 A full summary of the construction of the variables used in this paper is provided in the appendix. 9

10 median scaled net income for firms in the same Fama-French 48 industry. 3 Following Bertrand and Mullainathan (2001) and Bushman, Dai, and Wang (2010), we estimate the industry component of stock returns, referred to as luck, as the fitted value from cross-sectional regressions using one-year lagged annual buy-and-hold returns of the sample firms on the corresponding median Fama-French 48 industry return. The idiosyncratic component of stock returns, referred to as skill, is then estimated as the residual value from this fitted estimate. Panel B summarizes characteristics of directors, boards, and outside block owners as of the fiscal year-end prior to when turnover is identified. The first eight rows of variables in the panel are director characteristics. Summary statistics for these variables are computed as the average (median) for all of the director-firm-year observations in the sample. For example, we compute director age as the average (median) age of each observation for the full sample, which is (61.0) years. The average director tenure is 7.34 years. Nearly one third of the directors in our sample hold more than one public directorship at a given point in time, and roughly 4% gain a new directorship at a public firm in the year prior. Just over 20% of the director-firm-year observations are individuals who also hold a position as a CEO of another firm. One in ten director-firm-year observations are female, while 17.63% hold three or more public directorships. On average, over one third of director-firm-year observations are co-opted, which we define, following Coles, Daniel and Naveen (2014), as directors with tenure less than that of the current CEO. Panel B of Table 1 also includes measures of governance for the firm-year observations in the sample. Fifty (fifteen) percent of firm-years have a current (former) CEO that is also holds the title of Chair of the Board. On average, over 22% of observations have an active blockholder 3 Firms report industry peer firms to investors for benchmarking purposes; thus, we focus on performance measures relative to industry, which is also standard. We also evaluate stock price performance relative to a value-weighted market index and our results are statistically and economically equivalent to those throughout the paper. 10

11 in a given firm-year. 4 Our average board has 8.65 directors, where 71.63% are outsiders. Panel C reports firm characteristics that are used as controls in our analyses. Roughly 11% of the firmyear observations exhibit CEO turnover in the prior fiscal year. The average firm in our sample is years old with a market capitalization of $3.9 billion, and the volatility of stock returns for the average firm is In Panel D, we compare lagged firm performance for subsamples of turnover and nonturnover firm-years. Turnover by non-executive directors occurs in roughly 36% of the firm-years. Overall, the results in Panel D suggest that firm-year observations with turnover by at least one director are associated with relatively poor firm performance in the prior year when compared to firm-year observations with no director turnover. The average industry-adjusted stock return prior to a turnover event is -2.16%, which is 1.25% lower than the average industry-adjusted stock return for a firm-year observation when there is no director turnover. Similar differences obtain when we compare industry-adjusted accounting returns, and the stock returns attributable to skill The Sensitivity of Director Turnover to Firm Performance 4.1 Director-level estimates Table 2 summarizes the results of logistic regressions modeling the likelihood that an individual director experiences turnover in a given firm-year as a function of firm performance and attributes of the director and the firm. 6 The regressions also incorporate year fixed effects to capture any unmodeled macroeconomic trends. We report coefficient p-values in parentheses and standardized coefficients in brackets. The standardized coefficient relates the modeled effect on 4 In keeping with the literature (e.g. Clifford and Lindsey (2014)), we define an active blockholder as an outside shareholder who files a 13D disclosing 5% stake with intentions to affect firm or management. 5 Bushman et al. (2010) obtain similar results for skill and luck returns in firm-years associated with CEO turnover. 6 These specifications largely follow those in the director turnover models summarized in Fahlenbrach et al. (2014). 11

12 the likelihood of director turnover for a one standard deviation in a continuous variable, or for a change from 0 to 1 for an indicator variable. Models 1-3 incorporate continuous measures of firm performance. In Model 1, the coefficient on industry-adjusted stock return is negative and statistically significant, where the standardized coefficient indicates that a one standard deviation decrease in industry-adjusted stock returns is associated with an increase in the likelihood of director turnover of 0.6%. In Model 2, the coefficient on industry-adjusted ROA is also negative and significant suggesting that a one standard deviation decrease in industry-adjusted ROA is associated with an increase in the probability of turnover of 3.2%. In both models, the sensitivity of director turnover to performance is economically significant given an unconditional rate of turnover of 7.6%. 7 Model 3 estimates the sensitivity of director turnover to the stock returns attributable to luck and skill. The coefficient associated with luck captures the sensitivity of turnover to the component of performance presumably beyond the control of the board. The coefficient on skill, however, relays the turnover-performance sensitivity that can be ascribed to decisions of the directors and managers of the firm. The results of the model indicate that director turnover is not significantly related to luck, but is negatively correlated with skill. A one standard deviation decrease in skill related performance increases the likelihood of turnover by 0.6%, suggesting that director turnover is uniquely associated with stock performance attributable to actions taken by management rather than to events outside of their control. 7 Yermack (2004) documents that a one standard deviation change in contemporaneous market-adjusted stock returns is associated with a 0.92% increase in the likelihood director turnover. However, the study does not find a relation between director turnover and lagged firm performance. Our results are not directly comparable given his sample period, emphasis on larger firms, and restriction that the directors in the sample have held the position for a maximum of five years. When we restrict our sample observations to include only directors holding the position for five years or less, we find that the marginal effect of lagged performance on turnover is significantly higher. 12

13 In Models 4-6 of Table 2 we estimate director turnover as a function of measures of lagged firm performance constructed as indicator variables equal to one if the sample firm-year performance falls in the lowest quartile for a given sample year. The economic inferences drawn from Models 1-3 of the table are unchanged in these specifications. For example, the results in Model 4 suggest that the likelihood of turnover for directors of firms in the lowest quartile of industry-adjusted stock performance is 1.1% higher than the likelihood of turnover for directors on boards of firms in the top three quartiles of performance. Similarly, a director at a firm in the lowest quartile of industry-adjusted ROA is 1.8% more likely to turnover relative to a director in the top three quartiles of accounting performance. When we sort annual stock returns into quartiles of luck and skill, it continues to be the case that director turnover is sensitive to only the skill component of stock price performance. In the regressions summarized in Table 2 we control for a variety of director characteristics in order to consider whether specific attributes, such as reputation and experience, can help identify which directors depart a firm and which do not. 8 As noted earlier, identifying a performance turnover sensitivity for corporate directors may be significantly more complex than a similar sensitivity for CEOs given the collective action of boards. We therefore preface this analysis with three alternative predictions on the relation between performance related turnover on corporate boards, and the relative characteristics of the directors that do and do not turnover from a particular board. 9 First, it may be very difficult for a firm outsider, or even an insider, to ascertain the relative skill of an individual director. In this case, directors with the least experience or reputation may 8 We also include firm-level controls associated with CEO turnover, firm size, age and return volatility in each model in Table 2. For brevity, we suppress the output associated with these variables for this director-level analysis, but report them in the board level analysis in Table 3 and in Section 4.2. The statistical and economic significance of these firm-level control variables are not different in the analyses performed at either the director or firm level. 9 We note that none of these predictions are mutually exclusive in sample, and therefore our results in this section can be interpreted as indicating whether, on average, the effects associated with director skill and reputation have a moderating effect on the performance turnover sensitivity. 13

14 be scapegoated after poor performance to placate investors while avoiding the negative consequences of turnover of highly qualified directors. Alternatively, given limited information and attention, investors may simply hold the most visible board members accountable for poor performance. If this is the case, we expect that key committee chairs, and the chairman of the board are most likely to leave the board following poor performance. Finally, it is possible that internal monitors such as a lead director or chair of the board, or external monitors such as an active institutional blockholder, may be able to identify individual director actions and attribute these actions to firm performance. 10 In the regressions in Table 2 we include two indicator variables to control for director age; near retirement age (65-71) and average mandatory retirement age of 72 (or older) as in Cline and Yore (2014). As expected, the likelihood of director turnover is positively correlated with these age variables. Other characteristics related to a director s experience and reputation are also significantly related to turnover. For example, directors that hold more than one directorship are less likely to turnover, consistent with the hypothesized value of additional directorships in terms of connections and reputation. On the one hand, this result is consistent with the notion that directors exiting the board have the least amount of experience or reputation and are offered as a sacrificial lamb as their departure would be the least costly. On the other hand, less experienced directors may be of lower quality and identified as the poor performers on the board. In addition, directors who gained additional directorships in the prior year are more likely to turnover, suggesting that their increased responsibilities and limited attention may force directors to limit directorships held. While existing studies find CEO experience is positively correlated with the 10 In a conversation with the authors, a lead director of a Fortune 500 corporation (who wished to remain anonymous) relayed that...everyone on the board knows which directors are not cutting it, and we try to ensure that these individuals do not continue on as directors of the company. 14

15 likelihood of gaining additional board seats, our results indicate that current CEO experience is uncorrelated with director turnover. All three committee membership indicator variables (audit, compensation, and nominating) are significantly negative in Table 2. Audit committee membership has the greatest economic significance suggesting that directors on this committee are 3.3% less likely to turnover relative to directors not on this committee. This result is inconsistent with the notion that directors with the most visibility are held accountable for poor performance. We also find that female directors are 0.2% less likely to turnover than male directors, which is consistent with their relative scarcity in the director labor market as well as the hypothesized benefits of board diversity for corporate governance (e.g. Adams and Ferreira, 2009). Our results appear consistent with both the scapegoating and performance attribution motivations for director turnover. Disentangling the two motivations is extremely difficult given that both are likely to explain some of the turnover decisions in sample, and that directors with less experience and reputation are also likely to be poorly performing directors. To provide some evidence for performance attribution, we examine the performance turnover sensitivity for directors in firms conditioned on a measure of industry homogeneity. Parrino (1997) concludes that the performance of corporate executives is easier to identify in homogeneous industries because monitors can more precisely filter out performance shocks that are common to the industry all firms in that industry. In keeping with this notion, he finds that the turnover-performance sensitivity is higher for CEOs in homogenous versus heterogeneous industries. By extension, we hypothesize that internal or external monitors may be better able to attribute performance to individual directors in more homogeneous industries. We investigate whether the turnoverperformance sensitivity for directors varies by industry homogeneity. In untabulated results, we 15

16 find that the turnover-performance sensitivity for directors does not vary by industry homogeneity, a result that is inconsistent with performance attribution. Overall, our results suggest that performance-based turnover for corporate directors is more prevalent for directors with less visibility and experience and lower reputations. While we can t discount the notion that internal and external monitors are able to attribute poor performance to the actions of individual directors, or evidence, on average, is consistent with the scapegoating of inexperienced directors following poor firm performance. In sum, results in Table 2 document a persistent and negative correlation between past firm performance and director turnover. These findings are consistent with the notion that directors are disciplined for poor performance; a conclusion further corroborated by the fact that turnoverperformance sensitivity obtains only for the idiosyncratic component of stock returns. 4.2 Board-level estimates Given our evidence at the director level, we also examine the relation between firm performance and turnover at the board level. If director turnover is indeed related to prior performance, we expect that this relation will also manifest in a higher proportion of directors leaving the firm following poor accounting and stock price performance. Table 3 reports OLS regressions modeling the percentage of directors that turnover in a given firm-year as a function of lagged performance and board and firm characteristics. 11 As a baseline, we note that the unconditional rate of turnover for the sample is 5.9% of the average board membership in a given year, which translates into approximately 0.51 directors per year. In Models 1-3, we incorporate the same continuous measures of performance as in Table 2. In Model 1 (2), the coefficient estimate on industry-adjusted stock return (ROA) is negative and statistically significant 11 For robustness, we estimate Poisson count regressions modeling the number of directors that turnover in a given firm-year and find statistically and economically similar results. 16

17 suggesting that as firm performance decreases, boards experience a higher proportion of director turnover the next year. For example, a one standard deviation decrease in ROA is associated with a 2.3% increase in the fraction of board seat turnover in the following fiscal year. Model 3 separates stock returns into luck and skill. Consistent with our findings in Table 2, we find that the negative relation between the fraction of board turnover and stock performance holds only for the skill component of returns. In Table 3, we also report the effects of firm-level characteristics on the proportion of director turnover in a given year. Consistent with prior literature, we find that directors are more likely to exit immediately following the departure of the firm s CEO. In addition, we find that larger firms have more stable boards, and that board turnover increases in the volatility of stock returns and the size of the board. The regressions also include controls for firm age and the proportion of outside directors, however, their coefficient estimates are not significantly different from zero in the regressions. 12 Models 4 6 of Table 3 evaluate the proportion of director turnover for firm-year observations in the lowest quartile of lagged performance. The coefficients in each specification suggest that relatively poor firm performance is correlated with a higher proportion of board turnover. For example, the coefficient on low industry-adjusted stock performance in Model 4 is positive and statistically significant; firms in the lowest quartile of stock performance experience 0.9% more turnover a year. This effect is economically significant given that the average proportional turnover is 5.9% per year. Boards in the lowest quartile of industry-adjusted ROA experience 1.5% more turnover a year (Model 5). The sensitivity of board turnover to performance is significant for the idiosyncratic (skill) portion of firm performance, but not luck. Boards in the 12 We replicate Table 3 with the following director-level variables averaged at the firm-year level: director age, gain new seat, hold additional seats, current CEO elsewhere, and female and find qualitatively similar results (untabulated). 17

18 lowest skill quartile experience 0.9% more turnover in a given year. The sign and significance of the coefficients on our firm and board controls are all consistent with Models 1-3 of Table Our findings in Table 3 suggest a negative and economically significant relation between performance and the proportion of directors that turnover in the following year. This turnoverperformance sensitivity is robust to various measures of returns, and obtains specifically for stock returns attributable to management skill. Collectively, Tables 2 and 3 provide evidence that directors and boards as a whole are disciplined through turnover following poor firm performance. 4.3 Alternative explanations for turnover performance sensitivity In this section we consider three alternative explanations for the negative relation between director turnover and past firm performance. One alternative explanation is that directors may voluntarily leave to preserve their remaining reputational capital. As noted in Fahlenbrach, et al. (2014), voluntary turnover by directors seeking to preserve reputation is more likely to occur prior to the realization of poor performance. Given the structure of our tests, it is likely too late for directors to detach themselves from attribution for poor firm performance. Nevertheless, directors anticipating further poor performance may seek to avoid further damage to their reputation. Under this explanation, we should observe continued poor performance when directors exit after poor firm performance. A second alternative explanation is that directors may voluntary leave a board seat in an attempt to disassociate themselves from the poorly performing firm thereby providing a costly and credible signal of their quality to the labor market. In this case, directors leaving boards following poor performance should be more likely to experience a net gain in future directorships relative to the directors of the same firm that stay on the board. Finally, Farrell and Whidbee (2000) find an increased likelihood of outside director turnover following forced CEO turnover. 13 To evaluate the robustness of these results we re-estimate models in Table 3 using firm fixed effects. The sign and significance of each of our performance-turnover results are unchanged in these alternative specifications. 18

19 It is conceivable, therefore, that our director turnover-performance sensitivity is driven largely by coincident CEO departures and are the byproduct of a wholesale change in the management of the firm. We perform two tests to examine whether reputation concerns can consistently explain the negative sensitivity of director turnover to lagged firm performance. In untabulated univariate tests, we examine firm performance in the year following the departure of one or more directors. Average future industry-adjusted stock performance after turnover is 7.9% versus -2.2% the year prior, suggesting that, if anything, director departures are followed by improved firm performance. When we consider the subsample of firms in the lowest quartile of prior year industry-adjusted stock price performance, we find that firms experiencing director turnover realize an average industry-adjusted return of 12.9%, which is higher than the average industry-adjusted return of 7.3% for firms experiencing no turnover. We also examine future performance in a multivariate setting, controlling for firm and director characteristics and focusing on firms with and without director turnover in the year following performance in the lowest quartile for a given sample year. We find that future industry-adjusted returns are 6.1% higher for the subsample of firms that experience director turnover relative to firms that do not. In sum, our results are inconsistent with the notion that directors systematically depart firms following poor performance because they anticipate further poor performance and preemptively leave to preserve reputational capital. To evaluate the signaling alternative, we compare the change in the number of directorships and the probability of gaining a new board seat over a three-year window following the year of the departure for directors that leave a given board relative to those that remain. In Models 1-3 (4-6) of Table 4 estimate OLS regressions of the change in number of public board seats (the likelihood that a director obtains a new public directorship) in the three years after turnover. All 19

20 specifications control for director characteristics in Table 2 and include firm-year fixed effects which allows for within firm-year variation in director turnover and individual characteristics to explain changes in a director s future board seats. As a benchmark, the average change in board seats over a three-year period is -0.04, while the unconditional probability of gaining a directorship is 21.5%. 14 Model 1 of Table 4 reveals a significant and negative relation between turnover and the change in the number of a director s future board seats, suggesting that turnover directors experience ex-post settling up in the labor market. This relation manifests for directors who experience turnover from firms in the lowest quartile of prior stock performance (Model 2), but not for those in the highest quartile (Model 3). These results are consistent with notion that directors who turnover following poor firm performance are penalized rather than rewarded in the director labor market. Models 4-6 yield a similar conclusion. Directors that experience turnover are 8.8% less likely to gain a future seat relative to the directors that stay on a board in a turnover year. Model 5 suggests that the negative relation between turnover and the likelihood of gaining future seats is most pronounced for directors that turnover following very poor performance. Overall, contrary to the signaling prediction, we find that directors who turnover from boards incur indirect costs after turnover in terms of a net loss in future directorships in other public firms; an effect mainly attributable to the directors that turnover following very poor firm performance. To consider whether our director turnover results are largely driven by CEO turnover, we examine whether there are differences in the turnover-performance sensitivity of directors when there is turnover by the CEO in the prior year. Table 5 summarizes regressions similar to those in 14 In order to utilize firm-year fixed effects, for each firm-year at least one director on the board must experience a change in future board seats or gain a new seats in the three years following when turnover is identified. Otherwise, all observations within a given firm-year are completely determined and excluded from these analyses. 20

21 Tables 2 and 3, and include interaction terms between a CEO turnover indicator and our lagged performance measures. This interaction term estimates any differences in the director turnoverperformance sensitivity between years with and without CEO turnover. Models 1-3 estimate the likelihood that a director turns over in a given firm-year as a function of firm performance and characteristics, CEO turnover/performance interactions, and director attributes. We utilize a linear probability model to ease the interpretation of the marginal effects of interaction terms. 15 For brevity we report only the coefficients for our low performance measures and their interactions with CEO turnover. All of the director and firm controls included are the same as in Tables 2 and 3 and have the same sign and significance. In Model 1, the coefficients on low stock return and its interaction with CEO turnover are positive and significant. Consistent with prior work, the turnover-performance sensitivity is almost twice as high for directors in years after CEO turnover. Our results, however, are not driven solely by coincident CEO turnover given that the coefficient on low stock return is positive and significant. Directors of firms in the lowest quartile of past performance, absent CEO turnover, are 1.0% more likely to turnover in the next year. Model 2 examines the sensitivity of turnover to lagged industry-adjusted ROA and yields identical conclusions to Model 1. Model 3 delineates between luck and skill in stock return performance. The coefficient on low luck and its interaction with CEO turnover are both insignificantly different from zero. The coefficient on low skill is positive and significant, however, its interaction term with CEO turnover is insignificant. This finding suggests that CEO turnover does not moderate the turnover-performance sensitivity for directors when performance includes only the idiosyncratic component of returns. 15 Cornelli, Kominek and Ljungqvist (2012) and Guo and Masulis (2015) also use a linear probability model to estimate CEO turnover and interpret interaction terms. Results are also robust to continuous performance measures. 21

22 Models 4-6 of Table 5 estimate OLS regression specifications modeling the fraction of directors that turnover in a given year and include interaction terms between CEO turnover and the lagged performance measures. We find similar results at the board level relative to those at the director-level. In sum, the results of Table 5 suggest that coincident CEO turnover increases the sensitivity of director turnover to firm performance. While these findings fit with the prior literature on coincident turnover between CEOs and a firm s board, we also find that a substantial sensitivity of turnover to performance persists for directors in the absence of CEO turnover. 4.4 Performance turnover sensitivity following SOX The enactment of SOX in 2002, as well as subsequent changes in exchange listing requirements, led to dramatic shifts in board characteristics. Guo and Masulis (2015) note that boards rapidly shifted their composition to include predominantly outside and unaffiliated directors. Linck et al. (2009) also suggest that this period was characterized by increased scrutiny over board decisions by capital market participants and the public as well as an increase in director workload and accountability. Kaplan and Minton (2012) as well as Peters and Wagner (2014) document that CEO turnover, and its sensitivity to firm performance increased during this time. In this section, we examine whether there are any secular changes in the sensitivity of turnover to firm performance for directors, particularly following the period associated with the enactment of SOX. Table 6 summarizes regression specifications similar to those in Tables 2. In each, we estimate a linear probability model of the likelihood that an individual director experiences turnover as a function of lagged firm performance,, an interaction term between a post-sox year indicator and our performance measures, as well as controls for director and firm characteristics. Following Linck et al. (2009), the post-sox period is defined as 2003 onward. We focus on our 22

23 binary measures of relative performance for ease of interpretation, although the economic and statistical significance reported here are mirrored in specifications that utilize continuous performance measures. The director and firm controls in these specifications are the same as those used in Table 2. The coefficient on low stock return and its interaction with the post-sox time period are both positive and significant in Model 1 and suggest that the sensitivity of director turnover to past firm performance is more than twice as strong post-sox. Results in Model 2 indicate that a relation between director turnover and accounting returns existed pre-sox, but is not significant post-sox. This result indicates that investors focus more scrutiny on stock performance post- SOX. Model 3 separates stock returns into luck and skill components. While the coefficient on low luck is positive and statistically significant pre-sox, it essentially disappears post-sox. The coefficient on low skill and its interaction term are positive and significant indicating that the turnover-performance sensitivity associated with skill is more than twice as strong in the post- SOX era. The results of Table 6 are consistent with the enactment of SOX leading to a more dynamic director labor market, and one that is increasingly tied to idiosyncratic firm performance. 5. Corporate Governance and Director Turnover Prior studies examining executive turnover have emphasized the impact of a variety of governance features on the sensitivity of turnover to performance. In our context, the association between director turnover and firm performance is of substantial interest because board characteristics themselves are recognized as governance features that can have a significant effect on the sensitivity of executive turnover to performance. For directors, however, it remains unclear 23

24 which internal or external governance features might impact director turnover generally, or more specifically the sensitivity of director turnover to firm performance. In this section, we examine whether governance features, proxy voting, and ISS recommendations for or against a director s election moderate the likelihood of director turnover and the sensitivity of turnover to performance. We consider the effect of three internal governance features commonly associated with increased agency conflicts and are therefore predicted to insulate directors from internal monitoring. The first is a measure of director co-option by a CEO which is defined as director-firm-years in which a director s tenure is less than that of the current CEO (e.g. Coles et al. (2014)). The second is board busyness, measured as the percentage of directors that hold three or more public directorships (e.g. Fich and Shivdasani (2006)). The third are measures of CEO duality in which the current (or former) CEO also holds the title of Board Chair. The literature also demonstrates that institutions and other large external blockholders can increase the likelihood of executive turnover. 16 To examine the role of external monitoring on director turnover, we consider the effect of active blockholders defined as an outside shareholder who files a 13D disclosing a minimum 5% equity stake with the intention to affect change in a firm or its management. 17 As with our other variables, we estimate the effect of an active blockholder observed in the fiscal year prior to a prospective turnover year. To establish the effects of governance on turnover, Table 7 details results similar to those in Tables 2 and 3. We restrict our sample to the 109,331 director-firm-year and 14,560 firm-year observations with available data on active blockholders. Models 1-3 detail logistic regressions estimating the likelihood of director turnover in a given firm-year as a function of lagged firm 16 See for example Denis, Denis and Sarin (1997); Huson, Parrino and Starks (2001); and Goyal and Park (2002). 17 A number of papers establish the effects of active blockholders on a variety of corporate policies and firm value. For example, see Brav, Jiang, Partnoy and Thomas (2008) and Klein and Zur (2009). 24