Why Do CEOs Survive Corporate Storms? Collusive Directors, Legal Jeopardy, and Costly Replacement

Transcription

1 Why Do CEOs Survive Corporate Storms? Collusive Directors, Legal Jeopardy, and Costly Replacement M.D. Beneish, C.D. Marshall, J. Yang* First Draft: 27 May 2011 This Draft: 07 October 2011 Abstract We re-examine why CEOs remain in power in over half of the firms that intentionally misreport earnings. The failure of directors to use CEO dismissal as a reputation-cleansing device is even more puzzling after SOX, as regulators adopt non-prosecution policies for firms that selfpolice. We find that CEOs are more likely to remain when conventionally independent directors and CEOs appear to collude. That is, when both benefit from selling as insiders before the delinquent accounting is revealed, and when directors ratify one or more value-destroying mergers during the restatement period. We find retention is less likely when the misreporting is more severe and directors fear greater penalties from owners, lenders, and the SEC. Our results are robust to controlling for explanations based on the cost of replacing CEOs, the effectiveness of corporate governance, and the use of CFOs as scapegoats that have been offered in prior research. Overall, our analysis of directors personal incentives increases the explanatory power of the retention decision by approximately 30%. It suggests collusive trading and merger ratification as additional (and observable) means of assessing the independence of outside directors. JEL classification: G31; G32; G34; M40. Keywords: CEO turnover, fraud, restatements, corporate governance, insider trading Preliminary and Incomplete, Comments Welcome *Kelley School of Business, Indiana University. We have benefited from comments and suggestions by Nandini Gupta, Charles Hadlock, Dasol Kim, Todd Milbourn, Brian Miller, Charles Trzcinka, Xiaoyun Yu, and participants in the Finance brown bag seminars at Indiana University and the 7 th Corporate Finance Conference at Washington University in St. Louis. We thank Peter Demerjian for generously providing us with data on managerial ability.

2 I. Introduction We investigate why over half of CEOs in 427 misreporting firms in retain their jobs for at least one year after the public discovery. This finding is puzzling for several reasons. First, we only study severe misreporting events: sample firms restate to correct fraud or irregularities in financial statements. Second, CEOs of misreporting firms have been shown to often avoid losses by selling their firm s stock before the delinquent accounting is discovered [e.g., Beneish 1999, Li and Zhang 2006; Leone and Liu 2010]. Third, despite arguments that dismissing the CEO can restore financial reporting credibility, reestablish organizational legitimacy, and avoid costly audit and litigation outcomes, prior research documents similar rates of CEO retention (e.g., Arthaud-Day, Certo, Dalton, and Dalton 2006; Hennes, Leone and Miller 2008; Leone and Liu 2010). Fourth, directors failure to dismiss the CEO is also puzzling, because regulators increasingly adopt policies of non-prosecution for firms that cooperate and self police (Arlen 2011). Although CEO turnover has long provided financial economists with a laboratory to test hypotheses about the effectiveness of governance, there is surprisingly little directly studying the decision to retain CEOs. Existing research shows misreporting CEOs are more likely to survive when board monitoring is ineffective (there are fewer independent directors on the board or the CEO is a founder), when the misreporting is less severe, and/or replacing the CEO is more costly (e.g., CEO is founder). We extend a literature that has focused on the agency problem between managers and shareholders by considering the role of independent directors personal incentives in the CEO employment decision (e.g., a potential agency problem between directors and shareholders). That is, we add to prior research that suggests conventionally independent directors 2

3 are unable to replace the CEO, by considering conditions that make those same directors unwilling to do so. We propose that CEO retention decisions occur for the direct benefit of independent directors and for the benefit of existing shareholders (and the indirect benefit of directors). In particular, independent directors are less likely to remove CEOs when their equity trading and their ratification decisions suggest collusion with CEOs. We view directors as collusive if, like the CEO, they personally benefit from selling their equity during the period over which earnings are misreported. We also examine whether independent directors implicit ratification of firms investing and financing decisions during the restatement period reflect collusion. This follows from Jensen (2005) who suggests that prior to accounting fraud firms engage in value-destroying acquisitions, over-investment, and excessive equity issuance to sustain overvaluation. We posit that independent directors prefer not to attract attention to their own selling or to their ratification of economically deficient investing and financing choices. Thus, even though dismissing the CEO could enhance shareholder value by restoring credibility, we propose that directors whose actions align with those of CEOs have weaker incentives to replace the CEO, because doing so is akin to an admission of guilt. We also expect independent directors to act in shareholders interests (and indirectly in their own). Directors incentives to reduce the expected cost of the misreporting on the wealth of shareholders suggests they are less likely to remove a CEO when it is costly to find a replacement, and more likely to remove a CEO when the misreporting is more severe and expected costs of litigation are higher. As in prior work, we posit that CEOs are less likely to be removed when replacing them is more costly, but assess the costliness of replacement incrementally to existing work. To explain, in addition to controlling for performance 3

4 (accounting and market-based measures) and whether the CEO is a founder, we consider a measure of within-industry relative performance over the CEO tenure that we develop, and propose a modified hypothesis for founders that resign from the CEO post but remain as board chairs. 1 Second, we assess the costliness of retaining the CEO by considering the severity of the misreporting, the threat of litigation (shareholders, debt holders, and regulators) and the fate of the CFO. We evaluate whether directors are more likely to remove CEOs when the plaintiff s bar files a class action lawsuit against the firm, when the firm has borrowed excessively, and when the SEC investigates the firm s questionable accounting. Empirical leverage for our analyses comes from three sources. First, we control for a number of corporate governance characteristics and measures of board independence based on where directors work (insiders, affiliated, interlocked), how busy they are, and what and whom they know. 2 Second, we are less likely to make the classification errors identified by Karpoff, Lee and Martin (2008a) because our analyses require us to identify CEOs starting from one year prior to the period in which accounting is questionable/fraudulent and ending two years subsequent to public discovery. Thus, we are less likely to miss the ousting of a CEO who is directly or indirectly responsible for its occurrence or to incorrectly attribute turnover to questionable accounting. Finally, we allow for the fact that some CEO-founder resignations appear to be window dressing exercises as the founder remains as chair or in another leadership position on the board. Our analyses reveal the following: 1 For a subset of our sample with available data, we also consider a measure of managerial ability proposed by Demerjian, Lev, and McVay (2009) that is independent of firm performance. That is, these authors estimate the innate ability of a manager as the rank of their efficiency in transforming corporate resources to revenue relative to their industry peers. 2 In addition to traditional measures, we also evaluate whether independent directors are more effective in making CEO replacement/retention decisions when there is less information asymmetry (e.g., Duchin, Matsusaka, and Ozbas 2010; Masulis and Mobbs 2011) and a lower degree of social connectedness to the CEO (e.g., Cohen, Frazzini, and Malloy 2008; Hwang and Kim 2009). 4

5 (1) In terms of retention frequencies, we find that (230 out of 427) 54% of CEOs associated with intentional misreporting retain their jobs for at least one year after the market discovers the misreporting. This retention rate we observe is in line with prior research. Further, this retention rate increases to 57% if we consider leadership rather than CEO changes. That is, in 14 firms, the founder gives up the CEO hat, but remains as Chairman of the Board, effectively retaining command over the strategy and management of the firm. (2) In terms of insider selling, we find that when independent directors emulate CEOs selling, they are more likely to retain the CEO. We estimate the effect of collusive trading to increase the probability of retention by 10.0%. (3) In terms of ratifying decisions, we find that directors are more likely to retain CEOs, when they have ratified a merger that resulted in a loss of shareholder wealth. We estimate the effect of ratifying such mergers to increase the probability of retention by 14.4%. However, we find no evidence that abnormal capital expenditures or abnormal equity issuance influence the likelihood of retention. (4) In terms of costs of replacement, the evidence is consistent with economic intuition: a good track record increases the likelihood a CEO is retained by 11.2%. We also find that CEO-founders are more likely to be retained than non-founder CEOs (60.4 vs percent) consistent with Parrino (1997) and Leone and Liu (2010). Given instances where the founder gives up the CEO hat but remains as board chair, we find that founder CEOs are more likely to remain as the head of the firm (69.8 percent) than previously thought. (5) In terms of costs of retaining the CEO, we find that directors are less likely to retain CEOs when the firm is the subject of 10b-5 litigation. Indeed, we find that litigation precedes the CEO dismissal by 35 days for the median firm. We estimate the effect of the initiation of litigation proceedings to decrease the probability of retention by 20.3%. We also find that abnormal borrowing decreases the probability of retention by 6.8%. It is likely that firms borrowing abnormal amounts of public debt fire CEOs as a way to assuage bondholders. These results are robust to considering the effect of sample selection: we estimate a Heckman selection model where restating firms are distinguished from the COMPUSTAT population using a model that captures risk, profitability, growth, and pricing characteristics. They are also robust to controlling for year and industry fixed effects. Our analyses also incorporate controls proposed in prior research. Consistent with prior research, we find that CEOs are more likely to retain their jobs in firms with less adverse 5

6 performance, less severe accounting violations, and governance characteristics indicating less effective internal monitoring mechanisms (CEO and Chairman duality, large boards, and low outside presence on the board). We are able to increase the explanatory power of the decision to retain by approximately 30% relative to a benchmark model that is based on firm characteristics and traditional corporate governance measures (from 16.9% to 21.9%). At a minimum, our findings suggest collusive trading and merger ratification as additional (and observable) means of assessing the independence of outside directors. The remainder of the paper appears in four sections. In section 2 we present our empirical predictions. Section 3 describes the data. Section 4 presents the results of our empirical tests, and Section 5 concludes. 2. Empirical Framework Agency theory proposes that the board of directors is the pinnacle of authority in corporations where ownership is separate from control (Fama and Jensen 1983). Expert boards can hire, fire, and reward top management. They consist of both insiders and outsiders, with outsiders performing tasks where agency problems are significant (compensation committee, audit committee). Presumably directors have reputational incentives to carry out these tasks; that is, directors have incentives to build good reputations so as to retain their current directorships and potentially increase the number of boards on which they serve. 3 Although a wave of financial scandals has raised questions as to whether boards are able to prevent managerial wrongdoing (U.S. Senate Committee on Governmental Affairs 2002), it is not 3 Existing evidence does point towards the existence of reputational and wealth loss effects. In particular, Agrawal, Jaffe, and Karpoff (1999) find no evidence of increased director turnover, Srinivasan (2005) documents outside director turnover at firms with restatements but not at other firms where the director also serves. Studying lawsuits, Helland (2006) finds that outside directors of a sued firm experience an increase in the number of other board seats held following a lawsuit, and Fich and Shivdasani (2007) document that outside directors do not face abnormal turnover on the board of the sued firm but experience a significant decline in other board seats held. 6

7 clear that cases of financial fraud and questionable accounting have significant reputational consequences on outside directors. This is because it is often difficult, even for internal or external auditors who have access to detailed financial information and business transactions, to uncover the fraud. Indeed, directors could lay the blame on CEOs, suggesting the latter controlled the flow of financial information to the board so as to prevent detection, and would emerge from an accounting crisis unscathed. One way in which a board can convey its determination to ensure future financial reporting quality is to remove CEOs and/or CFOs either because these managers were involved in the questionable accounting or because they failed to effectively oversee the firm s accounting practices. Given large losses incurred by shareholders, how do directors justify retaining the CEO? 4 This study explores how tradeoffs from directors reputational cost perspective and form the firm s economic perspective affect the decision to retain the CEO for at least one year after the public discovery of an accounting irregularity. 2.1 Directors Collusive Actions First, we consider the role of directors equity incentives by examining whether independent directors sell their equity contingent wealth during the period over which earnings are misreported. There is evidence that CEOs abnormally sell their equity contingent wealth during periods in which earnings are overstated, and profit by doing so in advance of the public discovery (e.g., Beneish 1999; Li and Zhang 2006; Leone and Liu 2010; Ravina and Sapienza 2010). We examine whether the trades of independent directors and CEOs are concordant. If 4 The revelation of restatements and fraud is associated with significantly adverse abnormal returns (Beneish 1999, - 20% over three days (-1, +1); Karpoff, et al. 2008b, -25% on the first trigger event and -51% across all events). In addition, firms that announce restatements due to irregularities lose between 15 and 25 percent of their value in three months (Badertscher, Collins and Lys. 2008) and six months (Hennes, Leone and Miller 2008) after the restatement becomes public. 7

8 independent directors emulate CEOs selling and benefit from the misstated accounting, firing the CEO may attract the attention of regulators or the plaintiff s bar to their selling transactions. 5 We view the CEO and independent directors trading as a form of collusion and, in alternative form, predict that: H 1. All else equal, the probability a CEO is retained for at least one year after the public discovers financial misreporting is higher if both the CEO and conventionally independent directors engage in insider selling over our testing period (defined below). Our testing period is the intersection of (i) the period beginning one year prior to the start of the misreporting and ending at the last misreporting date, and (ii) the tenure of the CEO that is accountable for the misreporting. 6 To test this hypothesis we consider several measures. We use insider trading data from Thomson Financial, to assess whether both the non executive directors (presumably independent) and the insiders (CEO and executive directors, or alternatively only the CEO) are net sellers in the period beginning one year before the start of the restatement period and ending at the end of the restatement period till its end.. If both parties are net sellers (in shares or value) we designate this behavior as collusive insider trading. We also consider an indicator variable that identifies firms where non executive directors are net sellers. Second, we examine investing and financing decisions made by the CEO during the restatement period that were ratified by the board of directors. Our intuition is that if such actions destroy shareholder value, their occurrence suggests a degree of connectedness between the CEO 5 Our reading of lawsuit filings suggests that outside directors are rarely sued, and only named in complaints when they bail out in periods of declining prices. This is consistent with Srinivasan (2005). 6 If a CEO leaves office in the 18-month period around the restatement announcement date (-6, +12), we treat the CEO as accountable and dismissed. In addition, because the 18-month period is arbitrary, we checked all CEOs terminated and resigned outside the (-6,+12) months window by reading 8-K filings or searching Factiva to see whether the dismissal was fraud related. In the affirmative, we also coded those CEOs as accountable and dismissed. In cases where multiple CEOs tenures overlap the restatement period, we designate as accountable the CEO that has the largest overlap and that has departed closest to the announcement date. 8

9 and directors which casts doubt on the independence of the latter. The actions we consider are: mergers, capital expenditures and the issuance of possibly overvalued equity. In terms of mergers, a large literature has documented that many acquisitions occur for the benefit of managers (Jensen 1986; Shleifer and Vishny 1988; Morck, Shleifer, and Vishny 1990; Harford and Li 2007) or because managers overestimate their ability to manage the target company and overpay for acquisitions (Roll 1986; Moeller, Schlingemann, and Stulz 2004). 7 Although hubris likely does not apply, the private benefits explanation plausibly applies as directors that ratify acquisitions stand to benefit from higher compensation (Certo, Dalton, Dalton and Lester 2007). For this reason, we view directors who ratify value-destroying mergers during the CEO s tenure as colluding. In alternative form, we predict that: H 2. All else equal, the probability a CEO is retained for at least one year after the public discovers financial misreporting is higher if value-destroying mergers occurred during our testing period. We test the second hypothesis by assessing the board ratified one or more value-destroying mergers over the intersection of the CEOs tenure and the restatement period. We identify acquisitions using SDC Platinum, and assess the bidder wealth effect at announcement using CRSP data to compute three-day announcement abnormal returns (from a market model). Regarding other investment decisions, Kedia and Philippon (2009) show that lowproductivity firms that manipulate their earnings hire and invest too much to mimic the behavior of high-productivity firms. We posit that overinvestment destroys shareholder value, such that 7 This body of research shows that the wealth impact of an acquisition depends on the characteristics of the target (e.g., public vs. private), of the transaction (e.g., tender offer vs. merger, form of payment), and of the bidder (size, industry). In particular, acquisitions with the following characteristics are more likely to destroy value: (1) mergers (as opposed to tender offers -- See Jensen and Ruback 1983; Bruner 2002; Moeller, Schlingemann, and Stulz 2004 for reviews), (2) public targets (as opposed to private targets -- Fuller, Netter, and Stegemoller 2002; Moeller, Schlingemann, and Stulz 2004), (3) stock acquisitions paid for with equity signal that the equity of the bidder is overvalued (Travlos 1987; Fuller, Netter, and Stegemoller 2002; Moeller, Schlingemann, and Stulz 2004), (4) large bidders (Moeller, Schlingemann, and Stulz 2004) and bidders making diversifying acquisitions unless the bidders operate in sin industries (Shleifer and Vishny 1988; Morck, Shleifer, and Vishny 1990; Beneish, Jansen, Lewis, and Stuart 2008). 9

10 directors consenting to overinvestment during the period covered by the misreporting are likely to have a greater degree of connectedness with the CEO, and thus are more likely to retain the CEO. In alternative form, we predict that: H 3. All else equal, the probability a CEO is retained for at least one year after the public discovers financial misreporting is higher when the firm engages in abnormal investing during our testing period. In terms of financing decisions, prior work argues that managers prefer to issue shares if they perceive their stock is overvalued. This research often interprets the evidence of negative returns associated with equity issuances as a signal that the stock is overvalued (see among others, Asquith and Mullins 1986; Mikkelson and Partch 1986; Ritter 1991; Spiess and Affleck-Graves 1995; Stein 2003). To the extent that overvalued equity culminates in misreporting (Jensen 2005), we consider whether CEOs who were more successful in trading overvalued equity for cash are more likely to retain their job. Directors and CEOs who ratify such equity issuances may, in aggregate, be acting in the interest of shareholders even though these actions transfer wealth from new to existing shareholders. This could indicate a higher degree of connectedness with the CEO, and a higher probability of CEO retention. H 4. All else equal, the probability a CEO is retained for at least one year after the public discovers financial misreporting is higher when the firm engages in abnormal equity issuance during our testing period. To test the third and fourth hypotheses, we captures abnormal investing as the investment in capital expenditures, R&D, and acquisitions above that of the median firm in the industry, and equity issuance above that of the median firm during the restatement period (Data for these measures are from COMPUSTAT). 2.2 Costs of Replacement 10

11 Researchers have long recognized features of CEOs that make replacing them costly. Thus, a CEO who delivers superior performance during his tenure is more costly for shareholders to replace (Parrino 1997, Kaplan and Minton 2008, Burks 2010). For example, when Hewlett Packard fired Mark Hurd (due to his close relationship with a former contractor ) in August 2010, the stock price tumbled 8% in the first trading day after the announcement. In alternative form, we predict that: H 5. All else equal, the probability a CEO is retained for at least one year after the public discovers financial misreporting is higher when the CEO performs well relative to industry peers. Similarly, CEOs who are also founders are either more entrenched and/or more valuable. This makes them more costly to replace and, as has been shown in prior work, they tend to be retained more often (Parrino 1997; Adams, Almeida, and Ferreira 2005; Leone and Liu 2010). However, if the founder-ceo also serves as the Chairman of the Board, he/she is perhaps more willing to resign from the CEO post if they are able to retain control of the firm as board chair. In alternative form, we predict that: H 6. All else equal, the probability a founder CEO is retained for at least one year after the public discovers financial misreporting is lower when the founder remains as Chairman of the Board. We consider three measures for the cost of replacing the CEO following the revelation of financial misrepresentation. The first proxy for the cost of replacement we consider is the CEO s track record. We create a dummy variable that indicates under-performance (returns one standard deviation below the median) relative to industry peers in the period before the restatement. A second measure of costly replacement is used in prior work: it is an indicator of whether the CEO is also a founder. Our contribution is to also consider whether the founder 11

12 CEO also holds and retains the dual role of chairman of the board. Data for founder and chairman duality are collected from BoardEx and from proxy filings. 2.3 Costs of Retention When shareholders file lawsuits against the firm, it is likely too costly for the firm to retain the CEO. We predict that directors are more likely to lay blame on the CEO and seek to restore credibility by firing the CFO, and if necessary, the CEO. In alternative form, we predict that: H 7. All else equal, the probability a CEO is retained for at least one year after the public discovers financial misreporting is lower when shareholders file a lawsuit against the firm. Similarly, if the firm engaged in excessive borrowing during the restatement period we expect debt holders to have suffered greater losses. We believe the act of excessive borrowing during the restatement period may make it more costly to retain the CEO, since debt holders are more diligent in disciplining managers which is likely to lead to their removal In alternative form: H 8. All else equal, the probability a CEO is retained for at least one year after the public discovers financial misreporting is lower when the firm engages in abnormal debt issuance over our testing period. In measuring costly retention of the CEO, litigation proceedings are gathered from Stanford s Securities Class Action Clearinghouse. Abnormal debt issuance (net debt issued in excess of the industry median) is gathered from COMPUSTAT as a second measure of costly retention. 2.4 Control Variables We control for various internal and external corporate governance measures that have been shown to affect CEO turnover in firms with poor performance. Prior research suggests that the 12

13 sensitivity of CEO replacement to poor performance is higher in firms with higher proportion of outside directors (Weisbach 1988; Jenter and Lewellen 2010), in firms with outside blockholders (Denis, Denis and Sarin 1997), high institutional shareholdings (Huson, Malatesta and Parrino 2004), and in firms with lower ownership stakes by officers and directors (Denis, Denis and Sarin 1997), smaller boards (Jenter and Lewellen 2010), more homogeneous industries (Parrino 1997), and after the passage of SOX (Kaplan and Minton 2008). We control for a list of similar factors in our analysis of the CEO retention decision. We also consider the extent to which CEOs and directors have tied social networks because such connectedness has been shown to result in less effective governance. For example, directors who are socially connected to CEOs tend to grant higher CEO compensation with lower pay for performance sensitivity, and replace poor-performing CEOs less frequently (Barnea and Guedj 2007; Hwang and Kim 2009; Liu 2008; Larcker, Richardson, Seary and Tuna 2010). 8 We predict that directors who are socially connected to the CEO are more likely to retain the CEO when accounting irregularities are disclosed. Finally, recent research suggests the importance of information asymmetry between insiders and outsiders as a determinant of monitoring effectiveness (Duchin, Matsusaka, and Ozbas 2010). We consider several information asymmetry proxies from prior work including, stock return variance, and analyst following. 3. Sample and Descriptive Statistics The primary sources for the accounting restatements consist of the combination of Accounting and Auditing Enforcement Releases (AAER), GAO, and Audit Analytics databases. 8 Using a sample of firms under SEC enforcement actions for financial fraud, Chidambaran, Kedia and Prabhala (2010) show that shared educational and non-business antecedents between CEOs and directors increase fraud probability, while shared business connections reduce fraud probability. 13

14 Stock return and accounting data are from CRSP and COMPUSTAT respectively and traditional measures of corporate governance are from Compact Disclosure and SEC Filings. 3.1 Restatement Sample We draw our sample from three sources of accounting restatements. The first source includes firms that are charged with GAAP violations by the Securities and Exchange Commission in Accounting and Auditing Enforcement Releases. The second source consists of the subset of restatements related to irregularities from the GAO database. We rely on Hennes, Leone and Miller (2008) to identify restatements due to accounting irregularities from the GAO database. The third source consists of the subset of restatements related to irregularities from Audit Analytics where we again implement the technology from Hennes, Leone, and Miller (2008) to identify irregularities. As such, all restatements in our sample involve intentional misstatements. Table 1 reveals that we identify 739 misstatements from the AAER, GAO, and Audit Analytics databases. We drop 173 firms that are not in COMPUSTAT, 120 firms with missing announcement returns in CRSP, and 19 firms that are not in the Thomson Financial Insider Trading database. Our final sample consists of 427 restatements of intentional misreporting over the period 1993 to CEO and CFO Retention We hand collect CEO and CFO retention data from annual reports and proxy filings with the SEC and from press releases found in Factiva. We consider CEO and CFO turnover from six months before to twelve months after the month of the restatement announcement. To determine who is accountable for the misreporting, we track all CEOs and CFOs from one year prior to the period in which the questionable/fraudulent begins until two years subsequent to the public discovery of the misreporting. Firms often initiate investigations into accounting irregularities 14

15 months before the restatement announcement is made. As a result, the firm may make a turnover decision before the restatement is made public. Identifying the correct CEO is important, and prior work has shown that focusing on the officers in place at the time of the announcement can lead to the erroneous conclusion that the board failed to terminate the CEO when in fact the previous CEO had been replaced before the restatement was made public (Karpoff, Lee, and Martin 2008a). In Panel B of Table 1, we report a 53.9% (37.2%) retention rate for CEOs (CFOs) up to one year period after the restatement announcement. These retention rates are in line with prior research. 9 Figure 1 shows the timeline for the typical restatement and CEO retention decision. The average restatement period covers 2.5 years. If the board decides to fire the CEO, the average departure date is 2.5 months following the public announcement of the accounting restatement. We assess directors trading and ratification decisions over the testing period, which we previously defined as the intersection of (i) the period beginning one year prior to the start of the misreporting and ending at the last misreporting date, and (ii) the tenure of the CEO that is accountable for the misreporting. 3.3 Descriptive Statistics We compare characteristics of the firms that retain the CEO to those of the firms that fire the CEO in Table 2. Compared to boards that fire the CEO, boards retaining the CEO are more likely (1) to collude on insider trading transactions (44.8% v. 34.5%); (2) to have non-executive 9 A large number of studies have examined turnover in firms whose executives have engaged in questionable accounting. In studies involving firms charged with accounting fraud by the SEC, CEO turnover rates range from 36% in Beneish (1999, Table 6) to 88% in Karpoff, Lee and Martin (2008a, Table 5). In studies involving restatements, CEO turnover rates range from 8% for innocuous restatements to 49% for accounting irregularities (Hennes, Leone and Miller 2008). In studies that examine restatements, the rates of turnover vary depending on the type of executive, the period in which the sample restatements occur, and the window over which turnover is assessed. For example, the evidence in Hennes, Leone and Miller (2008) is based on a sample of restatements , and turnover is assessed in a 13-month period surrounding the restatement announcement. One year after the restatement, Land (2006) estimates that 45% of firms restating between 1996 and 1999 have CEO turnover; two years after the restatement, Desai, Hogan and Wilkins (2006) find 51% of restating firms in have turnover of their CEO, Chairman, or President, and Arthaud-Day, Certo, Dalton and Dalton (2006) observe CEO turnover in 43% and CFO turnover in 55% of their restatements. 15

16 directors who are selling equity (66.0% v. 56.9%), (3) to engage in value destroying mergers (47.8% v. 33.0%), and (4) less likely to issue stock excessively (10.8% v. 16.9%). Regarding costly replacement and retention measures, 18.7% of firms that retain the CEO had poor relative performance prior to the restatement compared to 33.7% of firms that fire the CEO. Among CEOs who retain their position 39.1% are founders, compared to 30.0% of founders among CEOs who get fired. Costly retention measures show that 44.4% of firms that keep the CEO are under litigation while 67.4% of firms that fire the CEO are under litigation. In addition, abnormal debt issuance is also associated with the CEO being more likely to be fired. Overall, univariate comparisons suggest that firms that retain the CEO exhibit a higher level of trading collusion between directors and CEOs, a greater incidence of value destroying mergers, and more significant social ties. We also observe that CEOs with a good track record (measured by prior performance relative to industry peers), and founder CEOs are more likely to retain their position. Firms involved in 10b-5 litigation proceedings, those with larger cash settlements from the lawsuit, and those with abnormal debt issuance are more likely to fire the CEO. The level of information asymmetry (measured by stock return volatility) is also higher for firms that fire the CEO. These univariate results suggest that the boards of restatement firms potentially have misaligned incentives and fail to remove CEOs because they do not want to draw attention to their own actions during the restatement period. On the other hand, boards seem to take into account the costs of replacing and retaining the CEO following the discovery of accounting irregularities, which may explain why some boards decide to retain the CEO even though the latter may have facilitated the misreporting. Further analysis is necessary to control for alternative explanations 16

17 offered in prior research including firm performance, the severity of misreporting, and conventional corporate governance measures. 3.4 Control Variables Firm characteristics for the sample of restatements with CEO retention and CEO firings are shown in Table 2 (Panel B). Confirming results of prior work, firms with more adverse accounting and stock performance and those with more severe restatements (e.g., SEC involvement) are more likely to fire the CEO. The three day cumulative abnormal announcement return is -7.4% for firms that keep the CEO compared to -14.8% for those that replace the CEO. The stock performance over the two-year period prior to the restatement announcement is -1.2% for the average firm that keeps the CEO compared to -36.5% for the average firm that fires the CEO. Accounting performance measured by cash flow to price is positive (0.0741) for firms that retain the CEO and negative ( ) for those that remove the CEO. In addition, the fraction of firms under SEC investigation (AAER) is 18.3% among firms that keep the CEO, compared to 49% among those that fire the CEO. Among firms that keep the CEO, only 45.2% of CFOs are removed, whereas if the CEO is fired, the CFO is also removed in 83.3% of the cases. We find more socially connected CEOs defined as having more social connections than the median firm-- are retained (27.8%) than fired (15.0%). 10 In terms of corporate governance characteristics, we find a higher proportion of independent directors in firms that retain the CEO this is surprising but the difference may not be economically meaningful (70% v. 65%). We also find that retaining firms have older boards and CEOs with longer tenures. 10 We measure the connectedness of the board as the degree of social connections between the CEO and other directors serving on the board at the time of the restatement announcement. Social connections are determined using BoardEx and are measured as any relationship between the CEO and other members of the board which occurs either during or prior to the restatement period. The connections can be through prior employment at the same firm, current or past external directorships, relationships developed while attending the same academic institution (Alumni Associations), or through membership in other social groups or clubs (Cohen, Frazzini, and Malloy 2008). 17

18 In Table 3, we compare our sample firms to the universe of firms in COMPUSTAT in terms of accounting performance and firm characteristics. We assign each firm in our restatement sample a percentile rank by comparing them to all firms in COMPUSTAT in the same year. The comparison shows that restating firms appear to be a slightly larger in size and have higher sales growth, but have significantly poorer accounting performance (measured by profit margin, earnings to price, and cash flow). 4. Empirical Tests We model the probability that a CEO is retained by the board of directors by considering the hypothesized proxies for directors incentives or their connectedness, the costs of replacement, and the costs of retention, while controlling for characteristics of the firm, its governance, and the nature of the questionable accounting: P (CEO Retained=1) = 1/(1+e -Y ), where Y= a 0 Directors Collusive Actions + a 1 Insider Trading Collusion + a 2 Value Destroying Merger + a 3 Abnormal Investing + a 4 Abnormal Equity Issuance Costs of Replacement + a 5 Under-Performance + a 6 Founder + a 7 Founder/Chairman Costs of Retention + a 8 Litigation + a 9 Abnormal Debt Issuance Firm Characteristics + a 10 Prior Stock Performance + a 11 Restatement Anncmt. Return + a 12 Cash Flow to Price + a 13 Bankruptcy Score + a 14 AAER + a 15 Stock Return Volatility + a 16 Analyst Coverage Corporate Governance Measures + a 17 CEO/Director Social Ties + a 18 CEO/Chairman Duality + a 19 Big Board + a 20 Board Independence + a 21 Old Board + error, 18

19 where: Insider Trading Collusion Value Destroying Merger Abnormal Investing Abnormal Equity Issuance Under-Performance Founder Founder/Chairman Litigation Abnormal Debt Issuance Prior Stock Performance = Indicator that both non-executive directors and insiders (CEO and executive directors) are net sellers during out testing period. = Indicator that there was at least one value destroying merger ratified by the board during our testing period = Firm s total abnormal investment that exceeds the median value for the corresponding measure in the firm s industry = Firm s abnormal net equity issuance that exceeds the median value for the corresponding measure in the firm s industry = Indicator that the firm's stock performance was one standard deviation below the industry median = Indicator that the CEO is the founder (or cofounder) of the firm = Indicator that the CEO is the founder (or cofounder) and also serves as the Chairman of the board of directors = Indicator that 10b-5 litigation (related to the corresponding accounting restatement) was filed = Firm s abnormal net debt issuance that exceeds the median value for the corresponding measure in the firm s industry = Market-adjusted monthly return from month -24 to -1 of the restatement announcement date Restatement Anncmt. Return = Market-adjusted three-day return from day -1 to +1 of the restatement announcement date Cash Flow to Price Bankruptcy Score AAER Stock Return Volatility Analyst Coverage =Cash flow from operations divided by market value of equity = Zmijewski s score (See Appendix). A higher score implies a higher level of financial distress = Indicator that the firm is subject to an accounting and auditing enforcement action by the SEC = Volatility of monthly market adjusted stock returns over a fiveyear period ending two years before the restatement announcement = Indicator when firm is followed by at least one analyst 19

20 CEO/Director Social Ties CEO/Chairman Duality Big Board Board Independence Old Board = Indicator that the percentage of social connections between the CEO and other current members of the board is greater than that of the median firm = Indicator that the CEO is also the Chairman of the Board = Indicator that the number of directors is greater than that of the median firm = Fraction of independent directors = Fraction of directors known to be over 69 years old Correlations between all variables used in the subsequent regression analyses can be found in Table 4. The correlation between the two alternative dependent variables is 93.6% suggesting that studying CEO or Leadership retention ought to yield similar results. Both of these variables are positively correlated with CFO retention, suggesting as we will later detail that CEO and CFOs are likely to be fired or retained in tandem. Generally speaking, the correlations between CEO retention (or Leadership retention) and various independent variables confirm the comparisons discussed in Table 2. For example, correlations with the insider trading collusion (.104 and.111) and merger variables (.150) are positive, whereas the correlation with litigation (-.232) and abnormal borrowing (-.098) are negative. There are also a number of correlations between independent variables that are large. Some are measures of the same underlying concept and are not used in the same regression e.g., variables [4], [5], and [6] on insider trading or variables [13] and [14] on litigation. Others independent variables measuring different attributes have correlations greater that.20 (in absolute value). As such, we examine variance inflation factors, and find no evidence that particular independent variables or groups of variables are inflating the variance of the parameter estimates. 20

21 4.1 CEO Retention Table 5 presents logistic regressions in which the dependent variable identifies CEOs that are retained for up to one year following the public discovery of the misreporting. The table contains four columns. Specification (1) contains twelve variables that are drawn from prior research and measure firm risk, performance, corporate governance, and the severity of the accounting violation. This specification is the baseline that contains control variables, and against which we measure the explanatory power of the tests of the hypotheses we propose. Consistent with prior work on CEO turnover, we find that firms with stronger accounting and stock performance are more likely to retain the CEO and that firms with less severe restatements in which the SEC was not involved are also more likely to retain the CEO. In Specification (2) we add several measures of directors collusive actions and in Specification (3) we add several measures of how costly it is to replace and to retain the CEO. Finally, in the last column we report the marginal effects for the variables based on the full model with 21 variables. We focus our discussion on the full model because results for variables that are common across specifications are similar. We find that the insider trading collusion dummy is positively related to the decision to keep the CEO. Because the measure indicates that both the CEO and independent directors were selling in advance of the discovery of the misreporting, this finding suggests that boards with apparently collusive insider trading behavior are more likely to retain the CEO following the revelation of the accounting irregularity because firing the CEO would attract attention to directors own trading. This evidence is consistent with H 1. It also corroborates findings in prior work that has shown that CEOs and directors profit by selling in advance of public discovery that earnings are overstated (Beneish 1999; Li and Zhang 2006; Ravina and Sapienza 2010). 21

22 We also find that boards that have ratified a value destroying merger during our testing period are more likely to retain the CEO. This is consistent with H 2. By contrast we find no evidence that the board s ratification of abnormal investing in capital expenditures (H 3 ) or abnormal issuance of equity (H 4 ) affect the probability of CEO retention. In terms of the cost of replacement, we find some support for the notion that better CEOs fare better: firms that performed worse than the industry median are less likely to retain the CEO (p-value=.059). This finding provides marginal support for H 5. Following prior research (Parrino 1997; Leone and Liu 2010), we control for cases where the CEO is also the founder and confirm that founder-ceos are more likely to remain as CEO. Furthermore, we separately examine cases in which a founder CEO is also the Chairman of the Board and find that being a founder and dual, reduces the likelihood of retaining the title of CEO. If the founder-ceo serves as the Chairman of the Board, he/she could be more willing to resign from the CEO post so long as the founder is able to retain control of the firm by remaining on the board. We find marginal support for this prediction (H 6 ) (p-value=.057). We also consider the cost of not replacing the CEO. We find that if the firm is subject to 10b-5 litigation, boards are less likely to keep the CEO (H 7 ). Indeed for the median firm, the CEO is dismissed approximately 35 days after litigation proceedings are announced. We find that boards that ratify abnormal amounts of debt financing are less likely to retain the CEO. We view this result as consistent with a higher cost of retaining the CEO given that excessive borrowing to sustain potential overvaluation means that a greater wealth transfer from bondholders to stockholders has occurred (H 8 ). The firing of the CEO is more likely to occur because bondholders are more diligent in monitoring and disciplining managers and directors. 22

23 In terms of the variables examined in prior work, we find that prior stock performance, restatement announcement returns, accounting performance and the issuance of AAERs by the SEC--all proxies for the severity of the misreporting--makes CEO retention less likely. We find no evidence suggesting that the extent of information asymmetry (stock return volatility, analyst following) plays a role in the likelihood of CEO retention. Our finding also suggest a limited role for traditional measures of the strength of governance (board size, board independence) and for more recent measures drawn developing literature that shows the role of social ties in assessing director independence. 11 Even though the univariate results revealed that socially connected boards were more likely to retain the CEO, the finding does not remain in the multivariate specification. Our model improves the explanatory power of the retention decision by approximately 30% relative to a benchmark model that considers only firm characteristics and conventional corporate governance measures (from 16.9% to 21.9%) in Table Leadership Retention A founder CEO who also serves as Chairman of the Board, may be more willing to resign from the CEO post if the founder fundamentally remains at the helm of the firm. We view these chairman-ceo-founder resignations as window-dressing exercises. They provide the appearance of effective board monitoring, but the dismissal of the CEO leaves the leadership of the firm unchanged. 11 We also consider the institutional origin of the social connection since prior work has found that non-professional connectedness may increase the probability of fraud whereas professional connections decrease the probability of fraud. (Chidambaran, Kedia, and Prabhala 2010 ) Perhaps in our sample the type of CEO/Director connection will affect the firing decision as well. Out of 359 CEO/Director connections only 18 were classified as non-professional connections meaning their origin was classified as academic, charity, foundation, or another organization. Since most of the connections that occur between the CEO and directors are professional (executives at another firm or directors on another board) our results do not change. 23

24 In Table 6, we present the results of the same specifications in the prior table, except that the dependent variable is equal to 1 if the CEO is retained, or if a CEO who is a founder and Chairman of the Board and remains on the board for at least six months after giving up the CEO title. With this new measure of Leadership retention many of our results are similar. We estimate that if there is insider trading collusion the CEO is 13.0% more likely to maintain a leadership position. If the board ratifies a value destroying merger the CEO is 14.3% more likely to retain a leadership role. CEOs with better track records are more likely to stay by 8.5%, but the finding is only significant at the 12% level. Those firms undergoing litigation proceedings are 21.9% less likely to keep the CEO at the firm. Our model improves the explanatory power of the retention decision by approximately 40% relative to a benchmark that considers only firm characteristics and conventional corporate governance measures (from 17.3% to 23.9%) in Table Robustness Tests In Table 7, we present the results using alternative measures for three of our variables of interest one at a time. In specification (1) we use litigation settlement data [a log transform] in lieu of the litigation indicator with similar results. In specification (2) we add measure of managerial ability proposed by Demerjian, Lev, and McVay (2009) that is independent of firm performance. As the indicator identifies managers with top quartile ability among industry peers, the positive coefficient is consistent with higher innate ability managers being more likely to keep their job; however, in this specification, the coefficient on the trading collusion variable becomes insignificant with a p-value of In specifications (3) and (4) we alter the definition of the insider trading collusion dummy. In specification (3) we measure trading collusion between CEOs and directors using the net value of insider sales, and in specification (4) we measure whether non-executive directors are net 24