The impact of accounting restatements on CFO turnover and bonus compensation: Does securities litigation matter?

Advances in Accounting, incorporating Advances in International Accounting 24 (2008) 162–171

Contents lists available at ScienceDirect

Advances in Accounting, incorporating Advances in International Accounting j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / a d i a c

The impact of accounting restatements on CFO turnover and bonus compensation: Does securities litigation matter? Denton Collins a,⁎,1, Austin L. Reitenga b,2, Juan Manuel Sanchez c,3 a b c

Rawls College of Business, Texas Tech University, Lubbock, TX 79409, United States Culverhouse College of Commerce, University of Alabama, Tuscaloosa, AL 35487-0220, United States Sam M. Walton College of Business, University of Arkansas, Fayetteville, AR 72701-1201, United States

a r t i c l e

i n f o

JEL classifications: M51 M52 Keywords: Earnings restatements Chief financial officers Executive compensation Executive turnover Contracting penalties, Disciplinary actions

a b s t r a c t This paper examines the association between accounting restatements, class-action securities litigation and chief financial officer (CFO) turnover and bonus compensation. We identify income-decreasing earnings restatements that were the result of aggressive accounting policies, and hypothesize that these restatements will result in higher CFO turnover rates, and lower bonus compensation, especially when the firm is the target of a restatement-related class-action securities lawsuit. Our results indicate that CFO turnover and bonus compensation are affected by restatements, but only when the restatement firm is the target of a classaction suit. When we expand the analyses to consider other types of executives (e.g., CEOs and COOs), we continue to find that turnover only occurs in the presence of a class-action suit. However, bonus compensation penalties to other types of executives are not limited to litigation-related restatements. © 2008 Elsevier Ltd. All rights reserved.

1. Introduction and synopsis This paper examines the association between accounting restatements, class-action securities litigation and chief financial officer (CFO) turnover and bonus compensation. Prior research has documented mixed evidence that top executives (e.g., chief executive officer, president, chairman of the board) of firms restating earnings are terminated because of the restatement event. Some published (or forthcoming) studies find that non-GAAP (i.e., not in accordance with generally accepted accounting principles) financial reporting is associated with increased top executive turnover (e.g., Arthaud-Day, Certo, Dalton, and Dalton, 2006; Desai, Hogan, and Wilkins, 2006; Collins, Masli, Reitenga, and Sanchez, in press), while other published work finds little evidence of increased turnover (Agrawal, Jaffe, and Karpoff, 1999; Beneish, 1999; Persons, 2006). Further, there is little published research on whether less severe punishments, such as reductions in compensation, are levied against top executives of restating firms. Arthaud-Day et al. (2006) and Wahlen (2004) point

⁎ Corresponding author. Tel.: +1 806 742 2098. E-mail addresses: [email protected] (D. Collins), [email protected] (A.L. Reitenga), [email protected] (J.M. Sanchez). 1 The authors thank their respective business schools for their financial support. Juan Manuel Sanchez also thanks the KPMG foundation for the financial support. 2 Tel.: +1 205 348 5780. 3 Tel.: +1 479 575 6113. 0882-6110/$ – see front matter © 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.adiac.2008.08.005

out that restatements of earnings that were previously overstated present researchers an interesting laboratory for examining the linkage between managerial competence/integrity/legitimacy and disciplinary actions, because the degree of the misstatement can be observed ex post. Public confidence in the ethical standards of business executives remains very low, and it is likely that confidence in firms taking disciplinary actions against managers engaging in fraudulent reporting is likely very low as well.4 Media coverage of corporate fraud and restatements from the late 1990s to date has been extensive, and a consensus seems to have developed in the popular press that executives go unpunished for earnings manipulations and even outright fraud (see, e.g., Lublin and Forelle, 2004).5 This perception has given rise to both recent legislation and litigation designed to

4 As of November 2005, Gallup News Service reports that only 16% of respondents rated “the honest and ethical standards” of forbes executives as very high or high (Jones, 2005). 5 For instance, an analysis carried out by USA TODAY suggested that some (but certainly not all) executives linked to five of the 10 largest earnings restatements in the US history (e.g., WorldCom, Rite Aid, Xerox, Cendant, Conseco, National Commerce Financial) have retained their jobs or they hold high-ranking positions with other important companies (Krantz, 2002). Of course, recent events have suggested that the criminal justice system is successfully prosecuting a number of these managers (see, e.g., Lublin and Rhoads, 2005), even when the managers claim to have been unaware of the financial improprieties at their firms. Thus, with the convictions and guilty pleas mounting, these perceptions of little or no punishment are likely to change over time.

D. Collins et al. / Advances in Accounting, incorporating Advances in International Accounting 24 (2008) 162–171

force managers to disgorge bonuses and other compensation that are tied to manipulated earnings.6 Such public perceptions, combined with the inability of prior research to establish a consensus on disciplinary actions, suggest that more research in this area is needed. A careful review of the existing literature identifies two areas that warrant further investigation. First, research into the restatementrelated turnover and compensation penalties levied against CEOs and CFOs is relatively limited. Second, there is little research examining the influence of class-action lawsuits on penalties imposed on executives following earnings restatements.7 Yet the ability of shareholders to bring such securities litigation is an essential component of corporate governance (Shleifer and Vishny, 1997). Absent stronger corporate governance, class-action litigation may be a mechanism that is effective in removing or levying punishment against incumbent management implicated in accounting restatements. Our research focuses on these two primary issues. First, following the call for research by Arthaud-Day et al. (2006), we extend the extant research to focus on both management turnover and bonus penalties, with our primary focus on the CFO of the firm. While termination is clearly a severe punishment for accounting misstatements resulting in restatement, other less severe penalties, such as cuts in bonus pay, can be imposed. We focus on the CFO due to her or his primary responsibility for the financial reporting process (Mian, 2001; Geiger and North, 2006), whereas, with the exception of Arthaud-Day et al. (2006) and Collins et al. (in press), most of the published research has focused on the CEO, president and chairman (e.g., Desai et al., 2006).8 In its passage of the Sarbanes–Oxley Act of 2002, the United States Congress saw fit to require CEOs and CFOs to personally certify the accuracy and completeness of the external financial reports of their respective firms (Geiger and Taylor, 2003). As such, under the law, the CEO and CFO are expected to share personal legal responsibility and to suffer similar consequences for fraudulent financial reporting. Implicit in this legal requirement is the notion that the CFO is in a position to influence significantly the financial results reported by the firm to investors and regulatory authorities (Geiger and North 2006). Finally, there is a pressing need for research into the impact of securities litigation as a means of imposing disciplinary action on firms engaging in aggressive financial reporting. While restatements are accompanied by an increased likelihood of securities litigation (Palmrose and Scholz, 2004), there is relatively little evidence suggesting that this external enforcement mechanism acts to impose discipline on managers when other governance mechanisms might fail, and little research examining the impact of classaction litigation in the restatement context.

6 With respect to legislation, Section 304 of the Sarbanes–Oxley Act of 2002 requires that, if a company is required to restate prior period earnings, the CEO and the CFO of the firm are required to reimburse the firm for any bonus or other incentive (equity) based compensation received by that person(s) during the twelve month period following the first public issuance or filing with the SEC (whichever occurs first) of the financial statements containing the overstatement. Section 304 also requires that the CEO and CFO of the company reimburse the company for any profits realized from the sale of securities of the firm during the 12-month period. With respect to litigation, a shareholder of Computer Associates International Inc. (CA) filed suit in late June of 2004 against 10 former and two current CA executives in order to force the executives to repay more than $1 billion in cash, stock options and restricted stock that the shareholder alleges were paid to the executives based on erroneous financial reporting (McDonald, 2005). 7 An unpublished working paper by Strahan (1998) reports that the likelihood of CEO turnover increases when a firm is subject to class-action litigation. Niehaus and Roth (1999) find that class-action litigation is associated with higher CEO turnover. However, neither paper examines the restatement context and neither focuses on CFOs. 8 Two concurrent working papers, Burks (2008) and Hennes, Leone and Miller (2007), have also focused on the CFO. Like Arthaud-Day et al. (2006) and Collins et al. (in press), they find evidence of restatement-related penalties levied against incumbent CFOs. However, none of these papers has examined the association between restatement-related class-action securities litigation and the imposition of penalties.

163

We contribute to the literature in this area in three significant ways. First, similar to Arthaud-Day et al. (2006), we extend research into restatement-related turnover for executives that are not at the highest level of the firm (i.e., officers other than CEO, president or chairman). The current round of scandals involving overstated earnings has implicated CFOs as well as CEOs.9 Given the shared responsibility of the CEO and CFO for financial reporting, it is important to understand the disciplinary actions taken against CFOs as well as CEOs. Second, we examine bonus compensation penalties to CFOs (and other executives) resulting from earnings restatements. This issue has not been addressed in the prior published literature, yet evidence on this issue relates directly to the debate accompanying Section 304 of the Sarbanes–Oxley Act. Third, we consider the influence of class-action lawsuits on the punishments imposed on CFOs. Prior research finds that restatements frequently trigger classaction and other lawsuits (Palmrose and Scholz, 2004), while other research suggests that lawsuits may be associated with lower executive compensation (Persons, 2006) and higher executive turnover (Strahan, 1998; Niehaus and Roth, 1999). Our research empirically links these two streams of research as they relate to the restatement context and directly tests the associations between disciplinary actions, accounting restatements and class-action securities litigation. The starting point for our sample of restatement firms is the General Accounting Office (GAO) report dated January 17, 2003 (GAO03-395R), which identifies firms that restated earnings between January 1, 1997 and June 30, 2002. We then add post-GAO report restatement firms identified by Baber, Kang, and Liang (2006) for the years 2003–2004. After reducing the original sample for various reasons (e.g., firms not in ExecuComp, restatements not the result of aggressive accounting leading to earnings overstatement, CFO data not disclosed, etc.), our final sample consists of 81 firms that overstated their earnings at least once in the period from 1997 to 2003, along with 81 control firms matched on the basis of industry, size and CFO data availability. Consistent with our predictions, we find that, in restatement years, CFO turnover rates are higher in the restatement firms relative to our control firms, and that CFO bonus compensation is lower in the restatement firms. However, our evidence suggests that CFOs are only penalized when the earnings restatement triggers a class-action lawsuit. Since it is possible that litigation only occurs when restatements are severe in nature, we re-estimate our models after including variables for the restatement amount scaled by assets, the litigation settlement amount, the issuance of an Accounting and Auditing Enforcement Release (AAER) and the market reaction to the restatement. We find that litigation — not common measures of restatement severity — is the determining factor in CFO penalties. This suggests that there is increased public pressure on the board to take action when the restatement is associated with a class-action suit (Palmrose and Scholz, 2004). Whatever the mechanism by which the litigation triggers disciplinary actions, the litigation-related restatement CFO turnover rate of 57% is extreme, even when compared to the CFO turnover rate in a separate sample of firms with extremely poor performance. When we expand the analyses to the two years before the restatement and the year after the restatement, we find no significant differences in turnover between the restatement firms and the control firms in the pre- and post-restatement period. However, we do find evidence of lower bonus payments to CFOs in the year before the restatement, which could be related to the firm having advance knowledge of the aggressive accounting problem. When we expand

9 Specifically, the Committee of Sponsoring Organizations of the Treadway Commission noted that, in the instances of fraudulent financial reporting it studied from 1987 to 1997, the CFO was implicated by the SEC in 43% of its Accounting and Auditing Enforcement Releases (AAERs; Beasley, Carcello, and Hermanson, 1999; see also Young, 2006).

164

D. Collins et al. / Advances in Accounting, incorporating Advances in International Accounting 24 (2008) 162–171

the analyses to include CEOs, top-level executives (CEO, president, chairman) and low-level executives (e.g., COOs), results are similar to the CFO results with one exception. We find some evidence of bonus penalties for executives in restatement firms that were not the target of a class-action suit. On the other hand, we find no evidence of higher executive turnover rates in restatement firms that were not the target of a class-action lawsuit, suggesting that restatement-related turnover is largely the result of shareholder litigation. The rest of our paper is organized as follows. We review prior related literature and develop our hypotheses in Section 2. Section 3 describes our sample selection procedure, Section 4 presents our research methodology, and Section 5 reports our results. We conclude the paper in Section 6 with a discussion of our results. 2. Literature review and hypothesis development 2.1. Prior research on the causes and effects of earnings restatements A number of prior studies have shown that restating firms underperform and are in worse financial condition than non-restating firms (e.g., Kinney and McDaniel, 1989; Defond and Jiambalvo, 1991; Desai et al., 2006; Collins et al., in press). Not surprisingly, the restatement event is associated with significantly negative abnormal returns (see, e.g., GAO-03-138, Palmrose, Richardson, and Scholz, 2004). Ironically, Richardson, Tuna, and Wu (2002) argue that firms involved in restatements managed their earnings in attempts to attract investor capital at lower cost. 2.2. Disciplinary actions accompanying earnings restatements 2.2.1. Employment termination Research examining termination penalties has recently been expanded beyond an earlier body of research. The earlier work includes Agrawal et al. (1999), Beneish (1999) and Persons (2006), who each find that fraud revelations have little disciplinary impact on top managers (defined as the CEO, President and Chairman). Agrawal et al. (1999) interpret this evidence as suggesting that linking existing top managers to fraudulent behavior occurring at lower levels of the organization could be problematic, and that replacing managers can be costly to the firm if the replacements are inferior to the incumbent managers.10 Beneish (1999) reports similar findings after looking at firms subject to SEC enforcement releases, while Persons (2006) finds that Wall Street Journal stories revealing frauds or lawsuits are not associated with greater top executive turnover. However, none of these papers directly examine the penalties for CFOs of restating firms. In contrast, evidence of increased restatement-related penalties is reported by Desai et al. (2006), Arthaud-Day et al. (2006), and Collins et al. (in press). Desai et al. examines penalties assessed against top managers (defined as the CEO, President and Chairman) and documents increased turnover resulting from the restatements. They extend their analysis to show that the displaced managers subsequently find it difficult to gain similar employment with other firms. Arthaud-Day et al. also look at penalties assessed against top managers, but they extend their analyses to focus on CFOs and members of the board of directors as well. Similar to Desai et al., they find that the restatement event is associated with higher CEO and CFO turnover, but they also report a greater likelihood of board of director and audit committee turnover. Collins et al. (in press) essentially replicate the Desai et al. research, except that they focus on the CFO, they examine the pre- and post-SOX periods, and they track the

10 Agrawal et al. do acknowledge that top managers may have lobbied their respective boards of directors to keep their jobs, and shifted blame to lower-level managers for the fraudulent actions. Such “scapegoating” would likely manifest itself in greater lower-level executive turnover that does not impact top management (Khanna and Poulsen, 1995).

subsequent employment outcomes of displaced executives. They report little difference in turnover rates between the pre- and postSOX periods, but they do report that terminated CFOs are less likely to find employment comparable to their positions at the restating firms.11 In summary, other than Arthaud-Day et al. (2006) and Collins et al. (in press), the published research on executive penalties has focused on the CEO, with mixed results as to the disciplinary actions taken. As such, relatively little is known about the factors that influence nonCEO turnover. As noted previously, we focus on CFOs because the CFO is primarily responsible for preparation of the financial statements. Since the restatements examined in this paper relate to, at best, aggressive accounting, or, at worst, fraudulent accounting, the CFO would likely bear responsibility for the decision to engage in questionable accounting practices. Further, such legal responsibility is now codified in the Sarbanes–Oxley Act of 2002, with CFOs subject to similar legal sanctions faced by CEOs for accounting misstatements. Thus, consistent with Arthaud-Day et al. (2006), our first hypothesis, in the alternative form, is as follows: Hypothesis 1. Ceteris paribus, restatement firm CFO turnover is higher than control firm CFO turnover. 2.2.2. Bonus reduction Termination is arguably the most severe disciplinary action that the firm can impose on an executive. It is possible that firms may choose to apply less severe penalties in some cases, and some researchers have speculated as much (e.g., Arthaud-Day et al., 2006). A less severe penalty could be a reduction in cash compensation, likely in the form of reduced bonus pay. In fact, under certain circumstances, this action is mandated in Section 304 of the Sarbanes–Oxley Act regardless of whether the manager is terminated. There is evidence that compensation committees intervene in the compensation process when earnings are accompanied by unusual or special items (Dechow, Huson, and Sloan, 1994; Gaver and Gaver, 1998; Duru, Iyengar, and Thevaranjan, 2002; Adut, Creedy, and Lopez, 2003), or when firms seek bankruptcy protection (Gilson and Vetsuypens, 1993). Similarly, restatements may be perceived as signals of poor managerial performance (Palmrose and Scholz, 2004; Desai et al., 2006) or lack of integrity (Wahlen, 2004), in which case compensation committees would be more likely to intervene and impose bonus penalties. There is some evidence to support such conjectures, at least for CEOs. Persons (2006) argues that revelations of “fraud/lawsuits” in The Wall Street Journal are associated with reduced top executive cash compensation. But Persons provides no insights into the penalties faced by CFOs of restating firms. For the reasons stated above, we expect that compensation committees will intervene and penalize CFOs for overstating prior-period earnings. Formally, our second hypothesis, in the alternative form, follows: Hypothesis 2. Ceteris paribus, restatement firm CFO bonus compensation is lower than control firm CFO bonus compensation. 2.2.3. Class-action securities litigation and disciplinary actions Palmrose and Scholz (2004) document that restatement firms and their managers are likely to be subjects of securities litigation, although such litigation is not automatic. Persons (2006) argues that revelations of fraud and/or lawsuits in The Wall Street Journal are associated with limited disciplinary actions against top executives, but

11 As noted previously, concurrent working papers by Burks (2008) and Hennes et al. (2007) examine CFO turnover pre- and post-SOX, while Burks (2008) also reports on compensation-related penalties associated with restatements. In contrast to our paper, Burks reports relatively little evidence that bonus payments are reduced as a result of restatements. Burks speculates that this result may be due to his exclusion of obvious frauds from his sample. But neither paper examines the influence of class-action litigation on disciplinary actions.

D. Collins et al. / Advances in Accounting, incorporating Advances in International Accounting 24 (2008) 162–171 Table 1 Sample reconciliation and distribution of restatements Number of restatement firms Panel A: Final sample reconciliation Total number of firms identified in the GAO reporta Less: firms not in ExecuComp Subtotal Less: Restatement increased income Restatement had no effect on income Company was accused by SEC but never restated Company was acquired by another company No evidence of restatement found Missing necessary data items Subtotal Plus: restatement firms identified in Baber et al. (2006) Subtotal Less: firms that do not provide CFO data Final sample of restatement firms

27 39 6 3 36 32 88 37 125 44 81

Panel B: Distribution of restatements Year 1997 1998 1999 2000 2001 2002 2003 Total restatement firms

3 10 10 5 19 14 20 81

845 614 231

a The identification of these firms begins by using a report released by the General Accounting Office on January 17, 2003 (GAO-03-395R). The report identifies 845 firms (919 restatements) restating earnings during the period from January 1, 1997 and June 30, 2002. Additional firms are identified from Baber et al. (2006).

does not directly test the impact of restatement-related class-action securities lawsuits. On the other hand, Strahan (1998) and Niehaus and Roth (1999) both report that the likelihood of CEO turnover increases when a firm is subject to class-action litigation. However, only one of these papers focuses on the restatement context (Palmrose and Scholz 2004) and none of the papers focus on CFOs. Our research empirically extends these limited findings as they pertain to CFO disciplinary actions. With respect to earnings restatements, the presence of a class-action suit could increase the likelihood or the severity of disciplinary actions for at least three reasons. First, the lawsuit could place more pressure on the board of directors to impose penalties on the CFO in order to deflect the negative publicity resulting from the restatement and the lawsuit and to reduce/mitigate any potential damages arising from any settlement or verdict (Agrawal et al., 1999; Arthaud-Day et al., 2006; Persons, 2006). Second, the litigation could be positively related to the egregiousness of the accounting irregularity and the decline in firm value resulting from the revelation of the accounting misstatement (Wahlen, 2004; Persons, 2006). Finally, class-action securities lawsuits could suggest even greater problems with management integrity or competence that accompany the accounting restatement (Palmrose and Scholz, 2004; Wahlen, 2004). Given this body of research findings, we therefore state our third hypothesis, in the alternative form, as follows: Hypothesis 3. Ceteris paribus, disciplinary actions against CFOs will be more severe when the restatement triggers a class-action securities lawsuit. 3. Sample selection and distribution of restatements We begin our sample selection using a report released by the General Accounting Office on January 17, 2003 (GAO-03-395R). The

165

report identifies 845 firms that restated earnings between the period January 1, 1997 and June 30, 2002. The first restriction we impose on our sample is that restating companies are available in ExecuComp, our source for executive compensation data. This restriction reduces the sample by 614 firms, leaving a total of 231 firms. Next, using 10-K/ As and/or press releases, we verify that the restatement is the result of an aggressive accounting policy that overstated earnings. To accomplish the validation, we follow a similar methodology as that employed by staff members of the GAO. We perform keyword searches in Lexis–Nexis and Proquest using words (and variations) such as “restatement” or “revisions” or “amendment” or “adjustment.” We also examine the SEC website and read 10-K/A reports to further the validation. Panel A of Table 1 provides a reconciliation of our sample selection from our initial universe of firms. We exclude a number of firms because of the characteristics of the restatement. Specifically, we lose 27 firms because the restatement increased income, 39 firms because the restatement had no effect on income, 36 firms for which we could find no external validation of any restatement, and 6 firms that were alleged by the SEC to have misstated earnings, but never actually restated earnings. Further, we exclude 3 firms that were acquired by other firms, and 32 firms that were missing essential data. After these initial screens, our sample is reduced to 88 firms that overstated their earnings in the period 1997–2002. We added 37 post-GAO report restatement firms identified by Baber et al. (2006), increasing the sample to 125 firms. Since we are interested in CFO penalties, we require that the restatement firm must disclose information about the CFO. This excludes 44 firms, leaving the final sample at 81 restatement firms from the time period 1997– 2003.12 Similar to both Arthaud-Day et al. (2006) and Desai et al. (2006), we use a matched-pair design. In other words, we match each restatement firm with a control firm based on size, two-digit SIC code, and CFO information/data availability. Executive compensation data are obtained from ExecuComp and financial variables are obtained from Compustat.13 Panel B presents the distribution of restatements in our sample by year. Since the GAO document begins with 1997 restatements, we have no restatements prior to 1997. 4. Methodology 4.1. Analysis of CFO turnover In order to determine whether CFO turnover is associated with earnings restatements and whether the association between CFO turnover and accounting restatements depends on whether a securities-related class-action is filed against firms, we estimate two logistic regression models:

CFOTURN1 : LASTYRi;t ¼ β0 þ β1 ROAi;t þ β2 ROAi;t−1 þ β3 RETURNi;t þ β 4 RETURNi;t−1 þ β 5 LNASSETi;t þ β 6 TRENDLASTYRi;t þ β 7 RESTATEi;t þ ei;t

12 We find no statistically significant differences between the 44 restatement firms without CFO data and the 81 restatement firms included in the sample. We conducted t-tests on: ROA, annual return, restatement amount scaled by assets, the log of CPIadjusted assets, three-day returns surrounding the announcement date, and the issuance of an AAER. 13 ExecuComp tracks compensation data for the top five executives (by title) of the firm for the period 1992–2006. It is important to note that ExecuComp aligns executive compensation (e.g., bonus payments) to the corresponding performance year. For instance, even though an executive may receive her/his bonus payment in a later period (say, three months after the fiscal year end, e.g., March 2003), ExecuComp aligns the bonus payment to the appropriate performance year (e.g., the year ending December 31, 2002).

166

D. Collins et al. / Advances in Accounting, incorporating Advances in International Accounting 24 (2008) 162–171

CFOTURN2 : LASTYRi;t ¼ γ0 þ γ1 ROAi;t þ γ 2 ROAi;t−1 þ γ3 RETURNi;t þ γ 4 RETURNi;t−1 þ γ 5 LNASSETi;t þ γ 6 TRENDLASTYRi;t þ γ 7 SUITi;t þ γ 8 NOSUITi;t þ τi;t The models are estimated using cross-sectional data from the 162 firms in the sample (81 restatement firms and 81 matched control firms). The dependent variable, LASTYRi,t, takes the value of 1 if year t is the last year in which the CFO for firm i is listed among the top executive officers, and 0 otherwise. ROAi,t and ROAi,t − 1 are the return on assets for firm i in years t and t − 1, winsorized at the 1st and 99th percentiles. RETURNi,t and RETURNi,t − 1 are the annual percentage change in stock price for firm i in years t and t − 1, also winsorized at the 1st and 99th percentiles. We include these four variables to control for the expected negative relation between firm performance and executive turnover (see, e.g., Chen, 2004; Carter and Lynch, 2004; Fee and Hadlock, 2004; Gilson, 1989). We control for size effects using the natural logarithm of CPI-adjusted total assets (LNASSET).14 Finally, we control for macro-economic trends in top executive turnover using the variable TRENDLASTYR. TRENDLASTYR is the median value, by year, for LASTYR using all ExecuComp executives, after excluding firms included in our sample. We expect TRENDLASTYR to be positively related to CFO turnover.15 Our variables of interest, RESTATEi,t in Model CFOTURN1, and SUITi,t and NOSUITi,t in Model CFOTURN2, are defined as indicator variables. Specifically, RESTATEi,t takes a value of 1 if firm i is a restatement firm in year t, and 0 if firm j is a control firm. Our Hypothesis 1 leads us to predict a significant, positive coefficient on RESTATE (i.e., β7 N 0); this would imply that the restatement event increases the likelihood of CFO turnover.16 On the other hand, in Model CFOTURN2, SUITi,t (NOSUITi,t) is coded one if the firm is a restatement firm and the firm was (was not) the target of a class-action suit, and 0 otherwise.17 Our Hypothesis 3 leads us to predict that the coefficient on SUIT will be positive and significantly greater than the coefficient on NOSUIT (i.e., γ7 N γ8). 4.2. Analysis of CFO bonus compensation To test whether there is a CFO bonus penalty associated with earnings restatements, and whether the association between CFO bonus penalties and accounting restatements depends on whether a securities-related class-action is filed against firms, we estimate the following OLS models for CFOs [with White (1980) standard errors]:

CFOBON1 : LNBONUSi;t ¼ λ0 þ λ1 ROAi;t þ λ2 RETURNi;t þ λ3 LNASSETi;t þ λ4 LASTYRi;t þ λ5 TRENDBONUSi;t þ λ6 RESTATEi;t þ νi;t

CFOBON2 : LNBONUSi;t ¼ f0 þ f1 ROAi;t þ f2 RETURNi;t þ f3 LNASSETi;t þ f4 LASTYRi;t þ f5 TRENDBONUSi;t þ f6 SUITi;t þ f7 NOSUITi;t þ ηi;t As with the CFOTURN models, the CFOBON models are estimated using cross-sectional data from the 162 firms in the sample (81 14 We include LNASSETS in our models to control for the possibility that litigation and non-litigation firms might differ along this dimension. However, we re-estimate our models without size in the model and our inferences are essentially unchanged. 15 Adding TRENDLASTYR as a control variable is similar in spirit to including year fixed-effects; however, TRENDLASTYR is more parsimonious because it does not use as many degrees of freedom. The same logic applies to the TRENDBONUS variable included in our bonus estimation. Estimating our models using year fixed-effects yields virtually the same results. 16 We verified that the CFOs that were terminated were in office at least one year prior to the restatement event. 17 We do not create a multiplicative interaction term using RESTATE and SUIT because none of the control firms (RESTATE = 0) was subject to a class-action securities lawsuit (SUIT = 0).

restatement firms and 81 matched control firms). The dependent variable LNBONUS is the CPI-adjusted natural log of bonus compensation. ROA, RETURN, LNASSETS, LASTYR and RESTATE are as previously defined in the CFOTURN models. We expect LNASSETS, ROA, RETURN and LNASSET to all be positively related to CFO bonus compensation. Adut et al. (2003) find that the overall market level of compensation explains a significant portion of firm-level compensation. Similar to the method used in Adut et al. (2003), we construct a trend variable, TRENDBONUS, by using the natural log of the CPI-adjusted median value of bonus compensation, by year, for all ExecuComp CFOs not included in our restatement sample. We expect TRENDBONUS to be positively related to CFO bonus compensation. We include LASTYR that is coded 1 if it is the CFO's last year. Since the last year the CFO is in office could be the result of forced resignations or terminations, it is possible that the CFO will also suffer a compensation penalty. However, it is also possible that forced CFO departures could be associated with increases in bonus compensation resulting from severance agreements. Therefore, we do not predict a sign for LASTYR. Finally, our variables of interest, RESTATE, SUIT and NOSUIT, are as defined previously; however, Hypothesis 2 leads us to predict a significant, negative coefficient on RESTATE (i.e., λ6 b 0), while Hypothesis 3 leads us to predict that the coefficient on SUIT will be negative and significantly less than the coefficient on NOSUIT (i.e., ζ6 b ζ7). 5. Results 5.1. Descriptive statistics and univariate analyses We next summarize the financial performance data, relative CFO termination frequency and CFO compensation data between restatement firms and control firms. To lend comparability to prior work and to consider the fact that CFOs and CEOs are part of the same executive team, we contrast CFOs with CEOs. Table 2 reports statistics unconditional on litigation status for both our sample firms and the control firms, while Table 3 reports the statistics for only our restatement sample firms conditioned on litigation status. Table 2 documents that mean ROA and the mean first lag of ROA are significantly lower in the restatement firms relative to our control firms. This is consistent with restatement firms being in relatively poor financial condition (Kinney and McDaniel, 1989; Defond and Jiambalvo, 1991). Surprisingly, mean RETURN and the mean first lag of RETURN are not significantly different between restatement firms and control firms, and we observe no significant difference in mean firm size between the two samples as well. For both CFOs and CEOs, mean LASTYR is significantly larger in restatement firms and mean

Table 2 Descriptive statistics 81 restatement firms versus 81 control firms matched by year, industry and CFO data availability Variable

81 restatement firms

81 control firms

t-statistic

ROAt ROAt − 1 RETURNt RETURNt − 1 ASSETSt LASTYRt (CFO) LNBONUSt (CFO) LASTYRt (CEO) LNBONUSt (CEO)

−0.026 0.013 0.036 −0.091 13,033.51 0.296 3.500 0.185 3.357

0.028 0.039 0.107 0.025 11,076.18 0.172 4.333 0.025 4.967

−2.56⁎⁎ −2.02⁎⁎ −0.82 −1.52 0.36 1.86⁎ −2.55⁎⁎ 3.43⁎⁎ −3.66⁎⁎

⁎⁎ (⁎) The mean value for the variable is significantly different between the restatement firms and the control firms at p b 0.05 (p b 0.10). ROA is the return on assets (winsorized at 1st and 99th percentiles), RETURN is the annual percentage change in stock price (winsorized at 1st and 99th percentiles), ASSETS is total assets (winsorized at 1st and 99th percentiles), LASTYR is coded 1 if it is the CFO/CEOs last year and 0 otherwise, and LNBONUS is the natural log of CFO/CEO CPIadjusted bonus compensation.

D. Collins et al. / Advances in Accounting, incorporating Advances in International Accounting 24 (2008) 162–171 Table 3 Descriptive statistics 81 restatement firms conditioned on class-action lawsuits Variable

51 non-lawsuit firms

30 lawsuit firms

t-statistic

RESTATEAMT ROAt ROAt − 1 RETURNt RETURNt − 1 ASSETSt LASTYRt (CFO) LNBONUSt (CFO) LASTYRt (CEO) LNBONUSt (CEO) CAR (−1,+ 1) CAR (−5,+5) AAER

0.048 0.008 0.025 0.121 −0.036 12,399.90 0.137 3.967 0.078 4.024 −0.033 −0.043 0.196

0.024 − 0.083 − 0.008 − 0.107 − 0.183 14,110.65 0.567 2.708 0.366 2.223 − 0.133 − 0.152 0.467

0.71 2.33⁎⁎ 1.48 1.79⁎ 1.33 −0.19 −4.53⁎⁎ 2.41⁎⁎ −3.41⁎⁎⁎ 2.65⁎⁎ 3.17⁎⁎ 3.15⁎⁎ −2.65⁎⁎

⁎⁎ (⁎) indicate that the mean value for the variable is significantly different between the restatement firms and the control firms at p b 0.05 (p b 0.10), based on two-tailed tests. RESTATEAMT = the amount of the restatement scaled by total assets (winsorized at 1st and 99th percentiles), ROA is the return on assets (winsorized at 1st and 99th percentiles), RETURN is the annual percentage change in stock price (winsorized at 1st and 99th percentiles), ASSETS is total assets (winsorized at 1st and 99th percentiles), LASTYR is coded 1 if it is the CFO/CEO's last year and 0 otherwise, LNBONUS is the natural log of CFO/CEO CPI-adjusted bonus compensation, CAR(− 1,+ 1) is the three-day abnormal return surrounding the restatement announcement (winsorized at 1st and 99th percentiles), CAR(−5,+5) is the eleven-day abnormal return surrounding the restatement (winsorized at 1st and 99th percentiles), and AAER = 1 if the restatement resulted in the issuance of an AAER and 0 otherwise.

LNBONUS is significantly smaller in restatement firms. This suggests that CFOs and CEOs are more likely to leave the firm when there is an earnings restatement, and that earnings restatements negatively influenced CFO and CEO bonus compensation. These univariate results clearly support our hypotheses 1 and 2. To identify which of our sample firms are involved in restatementrelated securities class-action lawsuits, we search the list of securitiesrelated class-action lawsuits recorded by the Stanford Law School Securities Class Action Clearinghouse.18 Specifically, we identify whether each of our sample firms are involved in class-action lawsuits that were the direct result of the earnings restatements examined in this paper. Of the 81 restatements examined in this paper, 30 (37%) were associated with class-action lawsuits. Table 3 reports the descriptive statistics for our 81 restatement firms conditioned on the presence of a class-action suit. The restatement firms facing class-action securities litigation have significantly lower ROA (but not lagged ROA), which is consistent with Palmrose and Scholz (2004). We also find that lawsuit firms have marginally lower annual returns (RETURNt) in the year of the restatement relative to non-lawsuit firms, and both the three-day mean abnormal return and eleven-day mean abnormal return surrounding the restatement announcement are lower for litigation firms (−0.13 versus −0.03, and −0.15 versus −0.04, respectively).19 We also find that litigation firms are more likely to receive an AAER than non-litigation firms. Finally, the amount of the restatement (scaled by total assets) and firm size do not differ between firms involved in litigation and firms not facing litigation. With respect to managerial consequences, the CFO turnover rate is significantly different between the two groups (p b 0.001). The CFO turnover rate for firms involved in class-action securities litigation is 57%, while the CFO turnover rate for non-litigant restatement firms (14%) is roughly equivalent to the CFO turnover

18 We find that litigation tends to be filed quickly. On average, 67% (80%) of the lawsuits are filed no later than one week (one month) after the announcement of the restatement. In contrast, AAERs are, on average, issued roughly 2.5 years after the announcement of the restatement. 19 The mean three-day and eleven-day abnormal returns for our sample of restating firms are −6.98% and − 8.37%, respectively. These amounts are in line with those documented in prior research (e.g., Palmrose et al., 2004; Collins et al., in press).

167

rate in the control firms (17%). Further, CFO bonus compensation is lower in the litigant restatement firms relative to the non-litigant restatement firms, although the difference is not as dramatic as the turnover results. The univariate results provide support for our Hypothesis 3, suggesting that penalties to CFOs stemming from earnings restatements are related to presence of class-action securities litigation. Comparing CFO turnover and CEO turnover in the litigation firms, we find that the rate of CFO turnover (57%) is higher than the rate of CEO turnover (37%). In untabulated results, we also find that CEO turnover increases the likelihood of CFO turnover and vice versa. In the 24 restatement firms with CFO turnover, the CEO turnover rate is 54%, while in the 15 CEO turnover firms, the CFO turnover rate is 87%. The results suggest that CFOs stand a very poor chance of remaining if the CEO departs, while CEOs have roughly a 50% chance of remaining if the CFO departs. 5.2. Multivariate analyses of CFO turnover and bonus compensation In Table 4, we estimate the two logit turnover models and the two OLS compensation models defined in our methodology section. Panel A reports the results for the turnover models CFOTURN1 and CFOTURN2, while Panel B presents the results for the compensation models CFOBON1 and CFOBON2. In the model CFOTURN1, the coefficient on RESTATE is positive (as expected) but not significant in the logit estimation, which is inconsistent with our Hypothesis 1. This is somewhat surprising, given the findings of Arthaud-Day et al. (2006), Collins et al. (in press) and Desai et al. (2006) that executive turnover is positively associated with restatements. To examine the possibility that restatement firm CFO turnover is influenced by litigation, we estimate the model CFOTURN2; the second column in Panel A reports these estimation results. The results for model CFOTURN2 are substantially stronger than the turnover model unconditional on litigation status. Specifically, our pseudo-R2 almost doubles when we replace RESTATE with SUIT and NOSUIT. Similar to model CFOTURN1, we continue to find that both RETURN and lagged ROA are reliably negative (one-tailed p b 0.05), which suggests there is a performance aspect to the turnover action. Specifically, we find that worse performance is associated with a higher likelihood of CFO turnover. However, the significantly positive coefficient on SUIT (one-tailed p b 0.01), combined with the insignificant result for NOSUIT, suggests that restatements are associated with higher CFO turnover only when class-action securities litigation is undertaken by shareholders. The test of whether β7 N β8 is significant at p b 0.001 (a one-tailed test), supporting our Hypothesis 3. This result is interesting. To explore the nature of the insignificant NOSUIT coefficient more closely, we revisit the univariate results by examining CFO turnover proportions between the control firms and the restatement firms that were not the target of a class-action suit. We find no significant difference between these two sets of firms. The univariate results also suggest that CFOs are only penalized when the earnings restatement results in a class-action suit. In Panel B of Table 4, we report the results of our estimations of the bonus compensation models CFOBON1 and CFOBON2 specified above. We report first our results for the model CFOBON1, with restatements unconditional on securities litigation status. With respect to our control variables, we find that CFO bonus compensation is positively related to firm size, firm return on assets, and macro-economic trends in CFO compensation. We find no association between LASTYR and bonus compensation, suggesting neither higher nor lower compensation for CFOs leaving in the year of the restatement. With respect to our Hypothesis 2, bonus compensation is significantly lower (onetailed p b 0.05) for restatement firms relative to control firms. The results clearly suggest that CFOs suffered a bonus compensation penalty as a result of the earnings restatement.

168

D. Collins et al. / Advances in Accounting, incorporating Advances in International Accounting 24 (2008) 162–171

Table 4 Analysis of CFO turnover and bonus compensation Panel A: Analysis of CFO turnover Variable Intercept ROAt ROAt − 1 RETURNt RETURNt − 1 LNASSETt LASTYR TRENDLASTYR TRENDBONUS RESTATE SUIT NOSUIT N Model χ2/Model F Pseudo R2/Adj. R2

Panel B: Analysis of CFO bonus compensation

CFOTURN1

CFOTURN2

CFOBON1

CFOBON2

Dependent variable: LASTYR −2.0204 (−1.16) −0.0096 (−0.63) −0.0423 (−1.70)⁎ −0.8694 (−1.97)⁎⁎ −0.3307 (− 0.73) −0.0787 (−0.64) – 7.2204 (0.76) – 0.4503 (1.11) – – 162 16.08⁎⁎ 0.091

Dependent variable: LASTYR −1.9576 (− 1.06) −0.0040 (−0.24) −0.0434 (−1.68)⁎ −0.7741 (−1.65)⁎ −0.1827 (−0.39) −0.0610 (−0.49) – 6.0535 (0.59) – – 1.4931 (2.96)⁎⁎⁎ −0.3983 (−0.76) 162 27.81⁎⁎⁎ 0.158

Dependent variable: LNBONUS −5.7442 (− 1.04) 0.0350 (2.92)⁎⁎⁎ – 0.6033 (2.04)⁎⁎ – 0.2554 (2.74)⁎⁎⁎ −0.1611 (−0.44) – 1.7545 (1.39) –0.6072 (− 1.99)⁎⁎ – – 162 7.64⁎⁎⁎ 0.198

Dependent variable: LNBONUS −5.6510 (−1.02) 0.0319 (2.64)⁎⁎⁎ – 0.5846 (1.99)⁎⁎ – 0.2552 (2.75)⁎⁎ 0.0383 (0.10) – 1.7292 (1.38) – −1.1430 (−2.58)⁎⁎ −0.3599 (−1.06) 162 7.02⁎⁎⁎ 0.207

⁎⁎⁎Significant at p b 0.01, ⁎⁎Significant at p b 0.05, ⁎Significant at p b 0.10. ROA is the return on assets (winsorized at 1st and 99th percentiles); RETURN is the annual percentage change in stock price (winsorized at 1st and 99th percentiles); LNASSET is the natural log of CPI-adjusted total assets; LASTYR is coded 1 if it is the CFO's last year, 0 otherwise; TRENDLASTYR is the mean value of LASTYR, by year, for all ExecuComp executives not included in the sample; TRENDBONUS is the natural log of the CPI-adjusted median value of bonus compensation, by year, for all ExecuComp executives not included in the sample; RESTATE is coded 1 if the firm is a restatement firm, 0 otherwise; SUIT (NOSUIT) is coded 1 if the firm is a restatement firm and (and not) the target of a class-action lawsuit, 0 otherwise. Reported p-values are based on two-tailed tests.

However, similar to our turnover models, after replacing RESTATE with SUIT and NOSUIT, we observe significant differences between litigant/non-litigant firms. Specifically, the coefficient on SUIT is significantly negative (two-tailed p b 0.01), while the coefficient on NOSUIT does not differ reliably from zero. Further, the test of whether ζ6 b ζ7 is significant at p = 0.049 (a one-tailed test). Together, these results support Hypothesis 3. In other words, CFOs receive lower bonuses only when the earnings restatement results in a class-action securities lawsuit. 5.3. Sensitivity/robustness tests 5.3.1. Pre- and post-restatement periods Similar to the analysis performed in Desai et al. (2006), we investigate the possibility that our matching procedure did not adequately control for idiosyncratic differences between the restatement firms and the control firms. Therefore, we look at turnover/ compensation disciplinary actions in the time periods before and after the earnings restatement year. Specifically, we examine the two years prior to the restatement year and the year after the restatement year. If the firm was aware of the accounting misstatement well before the restatement, it is possible that the subsequent restatement could influence disciplinary actions in the year before the restatement year.

On the other hand, we believe that it is unlikely that the restatement would have any influence two years before the restatement year. Consistent with Desai et al., it is also possible that the effect of the restatement could carry forward into the following year. However, we should find the strongest evidence of penalties in the restatement year. In Table 5, we report the results of estimating the turnover model (CFOTURN) and the bonus model (CFOBON) for year −2, year −1 and year +1 relative to the restatement year (year 0). Since we only find evidence of penalties when there is class-action securities litigation, we estimate and report the results for models CFOTURN2 and CFOBON2. For ease of exposition, we include only the coefficient estimates for SUIT and NOSUIT. In the years surrounding the restatement event, we lose some matched pairs of firms due to data not being available. When examining the years surrounding the restatement year, we find only one instance where the restatement firms are significantly different from the control firms. Bonus compensation is significantly lower in the year before the restatement for CFOs in restatement firms facing class-action securities lawsuits. This could indicate that the firm was aware of the accounting irregularity before the official earnings restatement announcement. There are no differences between the restatement firms and the control firms two years before the

Table 5 Analysis of CFO turnover and bonus compensation in the years surrounding the restatement year Model CFOTURN2 dependent variable = LASTYR

Model CFOBON2 dependent variable = LNBONUS

Coefficient (z-statistic)

Coefficient (t-statistic)

Variable

Year − 2

Year − 1

Year + 1

Year − 2

Year − 1

Year + 1

SUIT NOSUIT N Model χ2/Model F Pseudo-R2/Adjusted-R2

0.210 (0.22) − 0.079 (−0.10) 114 8.59 0.119

0.185 (0.19) 0.386 (0.53) 144 9.84 0.127

1.121 (1.52) 0.808 (1.26) 92 17.54 0.173

0.752 (1.55) 0.382 (1.03) 114 2.59 0.090

− 0.973 (−2.26)⁎⁎ − 0.511 (− 1.56) 144 3.56 0.111

−0.655 (− 1.14) −0.490 (− 1.05) 92 4.03 0.189

Intercept and control variables are excluded from the table. ⁎⁎Significant at p b 0.05, ⁎Significant at p b 0.10. LASTYR is coded 1 if it is the CFO's last year and 0 otherwise, LNBONUS the natural log of CFO CPI-adjusted bonus compensation, Year − 2 is 2 years before the restatement year, Year − 1 is one year before the restatement year, Year + 1 is one year after the restatement year, SUIT is coded 1 if it is a restatement firm and the target of a class-action lawsuit suit, and NOSUIT is coded 1 if it is a restatement firm and not the target of a class-action suit. For model CFOTURN2 and CFOBON2 control variables, see Table 4. Reported p-values are based on twotailed tests.

D. Collins et al. / Advances in Accounting, incorporating Advances in International Accounting 24 (2008) 162–171

restatement and one year after the restatement, which suggests that there are not idiosyncratic differences between the restatement firms and the control firms. 5.3.2. CEO turnover and CFO turnover Next, we control for the possibility that CEO terminations may be explaining the abnormally high rate of CFO departures in restatement years. Fee and Hadlock (2004) find that CEO turnover increases the likelihood of non-CEO turnover. Our univariate results reported in Section 5 similarly suggested that CEO and CFO termination are strongly related to each other. This could suggest that the higher turnover rate for CFOs in restatement firms facing class-action securities litigation is simply a function of the higher turnover rate for CEOs in those firms. To examine this possibility, we removed all CEO turnover firms (along with the matched firm) from the sample. When we re-estimate the turnover model CFOTURN2 using this subsample of firms (n = 126), the coefficient on SUIT continues to be positive and significant (p = 0.026, one-tailed). We also estimated the CFO turnover model after including an indicator variable for CEO turnover and found similar results. These results indicate that restatement-related litigation is associated with higher CFO turnover, even when the CEO remains in office, which suggests that the impact of restatement-related litigation on CFO turnover is incremental to the effect of executive teams being terminated as a group. 5.3.3. Magnitude of CFO turnover rate We next examine the extent to which the magnitude of CFO turnover rates in restatement firms facing class-action securities lawsuits is unusual. To examine how extreme the CFO turnover rates are for these restatements, we draw a sample of extremely poor performing firms from ExecuComp, and compare the CFO turnover rates for these poorly performing firms with the CFO turnover rates for our restatement firms facing class-action lawsuits. We identified 2920 firm-years in ExecuComp where both ROA and RETURN were in the bottom decile of the ExecuComp distribution. These firm-years are characterized by extremely poor performance, and presumably high executive turnover rates. The CFO turnover rate in these firms is 31%, which is considerably higher than the 17% CFO turnover rate in the control firms. We next performed a t-test comparing the CFO turnover rate for the 30 class-action litigant restatement firms (57%) to the CFO turnover rate in the 2,920 poor performance firm-years (31%). The difference is significant at p = 0.002. In other words, the CFO turnover rate observed in the litigant restatement firms is extraordinarily high, even when compared to CFO turnover rates in extremely poorly performing firms. 5.3.4. Non-CFO disciplinary actions We next examine the extent to which our CFO results generalize to other classes of executives. Thus far, our analyses suggest that only restatements that generate class-action securities lawsuits result in disciplinary actions being taken against CFOs. This suggests that we focus on turnover/bonus disciplinary actions against other classes of executives, after controlling for the presence of a class-action suit. We identify three groups of executives for analysis: (1) CEOs, (2) top executives (i.e., all individuals with the title CEO, President or Chairman or COO), and (3) lower-level executives, which includes all executives that do not have the title CEO, President, Chairman, CFO or COO. As in Desai et al. (2006), we define LASTYR for top executives to equal 1 if there is any turnover among that group of executives, and zero otherwise. We use the same procedure for lower-level executives. For bonus compensation, in order to have one observation per firmyear, we use the approach employed by Carter, Lynch, and Tuna (2004). Specifically, we define LNBONUS to be the mean value for bonus compensation, per year, per firm, for each group of executives. We re-estimate models CFOTURN2 and CFOBON2 used in Tables 4 and

169

5, except that we use the revised definitions of LASTYR and LNBONUS for the three groups of executives. We focus on the following four time periods: 1) two years before the restatement, 2) one year before the restatement, 3) the restatement year, and 4) one year after the restatement. Our results are reported in Table 6. Results are generally consistent with those already reported for CFOs, with two significant exceptions.20 First, we find evidence of CEO turnover in the year prior to the restatement. Again, this could be due to the firm having advance knowledge of the restatement and removing the CEO in order to frame the restatement announcement to the public. But it could also be due to the CEO leaving the firm because s/he (1) knew of the irregularity and (2) exited the firm prior to the irregularity being discovered. It is unclear how these two explanations could be disentangled. Second, we find evidence that, in the year of the restatement, non-litigant restatement firms appear to impose bonus penalties on all classes of executives other than the CFO, although the penalties are significantly smaller (p b 0.06, one-tailed), and roughly half those associated with litigant restatement firms. Recall that Table 4 reported no significant association between bonus compensation and non-litigant firm status for CFOs. The reason for this difference is unclear and is open for conjecture. We continue to find no differences between restatement firms and control firms two years before the restatement and one year after the restatement. Finally, we continue to find no evidence of executive turnover when the restatement did not lead to a class-action suit. 5.3.5. A closer look at litigation We find that class-action litigation is strongly associated with disciplinary actions following earnings restatements. Class-action lawsuits are also associated with other restatement-related factors. For example, in Table 3, we find that restatement firms with suits had more negative returns surrounding the restatement and were more likely to receive an AAER. It is also possible that the outcome of the suit may influence disciplinary actions. In an effort to unwind these intertwined factors, we re-estimate the models presented in Table 4 after including three-day returns surrounding the announcement date, the issuance of an AAER, the restatement amount scaled by assets and the settlement amount of the litigation in the models.21 For both CFOs and CEOs, results are qualitatively unchanged from those reported in the paper when these additional restatement-related explanatory variables are included in the model. We interpret this to mean that the influence of litigation on disciplinary actions is separate from other measures of the magnitude or egregiousness of the restatement. If the litigation itself, rather than other factors associated with the litigation, influences disciplinary action, suits occurring quickly should have more influence on disciplinary actions. To examine this issue, we partition the suits between those occurring within 30 days of the announcement and those that were delayed.22 CFO (CEO) turnover is 66.67% (41.67%) in the 24 restatement firms where the suit occurred quickly versus 16.67% (16.67%) in the 6 restatement firms where the suit is delayed. Taken together, the analyses suggest that the litigation event has a direct influence of disciplinary action rather than litigation being a proxy for an unidentified measure of restatement magnitude.

20 We also estimated the CEO turnover model after including an indicator variable for CFO turnover and continued to find a significant relationship between restatement litigation and CEO turnover. 21 We find qualitatively similar results using a five-day window surrounding the restatement announcement. 22 We find qualitatively similar results if we classify suits occurring within 90 days of the announcement as being “quick.”

170

D. Collins et al. / Advances in Accounting, incorporating Advances in International Accounting 24 (2008) 162–171

Table 6 Analysis of Non-CFO turnover and bonus compensation the years surrounding the restatement year Model CFOTURN2 dependent variable = LASTYR

Model CFOBON2 dependent variable = LNBONUS

Coefficient (z-statistic)

Coefficient (t-statistic)

Year − 2

Year − 1

Year 0

Year + 1

Year − 2

Year − 1

Year 0

Year + 1

CEO SUIT NOSUIT N χ2/F R2

0.330 (0.24) −0.241 (−0.19) 148 10.59 0.242

2.195 (2.42)⁎⁎ 0.948 (1.02) 156 15.28 0.181

3.258 (3.41)⁎⁎ 1.013 (1.05) 162 45.13 0.424

0.408 (0.50) − 0.287 (−0.39) 112 19.14 0.208

0.017 (0.03) 0.251 (0.57) 148 4.11 0.129

−0.807 (−1.41) −0.538 (−1.21) 156 5.06 0.155

−2.264 (−3.66)⁎⁎ −0.988 (−2.14)⁎⁎ 162 7.53 0.221

−0.681 (−1.10) −0.657 (−1.26) 112 5.48 0.220

TOP SUIT NOSUIT N χ2/F R2

0.441 (0.55) −0.627 (−0.78) 148 10.22 0.123

0.231 (0.41) −0.649 (−1.19) 156 6.44 0.045

1.221 (2.11)⁎⁎ 0.599 (1.10) 162 21.12 0.142

− 0.167 (−0.23) − 0.464 (−0.74) 112 19.62 0.173

0.129 (0.25) 0.247 (0.61) 148 4.62 0.147

−1.133 (−2.16)⁎⁎ −0.448 (− 1.08) 156 4.99 0.153

−1.818 (−3.19)⁎⁎ −0.858 (−1.92)⁎ 162 7.14 0.211

−0.540 (−0.91) −0.631 (−1.25) 112 5.30 0.213

LOWER SUIT NOSUIT N χ2/F R2

0.575 (1.17) −0.050 (−0.13) 144 7.16 0.037

0.748 (1.57) 0.507 (1.36) 154 4.60 0.022

1.102 (2.05)⁎⁎ 0.310 (0.83) 156 14.90 0.070

0.781 (1.25) 0.424 (0.90) 110 18.36 0.121

0.105 (0.29) 0.058 (0.21) 144 7.51 0.242

−0.601 (−1.79)⁎ −0.059 (−0.22) 154 8.07 0.244

−1.329 (−3.46)⁎⁎ −0.571 (−1.91)⁎ 156 7.49 0.227

−0.475 (−1.12) −0.511 (−1.46) 110 4.73 0.193

Intercept and control variables are excluded from the table. ⁎⁎Significant at p b 0.05, ⁎Significant at p b 0.10. See Section 4 of the paper for our definitions of LASTYR and LNBONUS. CEO is a subsample consisting of CEO observations, TOP is a subsample consisting of top-level executives (CEO, President, Chairman or COO), LOWER is a subsample consisting of lower-level executives, Year − 2 is 2 years before the restatement year, Year − 1 if one year before the restatement year, Year 0 if the restatement year, Year + 1 is one year after the restatement year, SUIT is coded 1 if it is a restatement firm and the target of a class-action lawsuit, and NOSUIT is coded 1 if it is a restatement firm and not the target of a class-action lawsuit. For model CFOTURN2 and CFOBON2 control variables, see Table 4. Reported p-values are based on two-tailed tests.

5.3.6. The influence of SOX Since SOX requires bonus disgorgement when bonus compensation was received as a result of overstated earnings, it is possible that our bonus results could be driven by post-SOX observations. To examine this possibility, we re-estimate the bonus model for CFOs and CEOs after partitioning the sample between pre-SOX (n = 94) and postSOX (n = 68) observations. In both subsamples, we find results generally consistent with those presented in the paper.

This paper contributes to the executive discipline research by demonstrating that CFOs suffer relatively severe penalties when aggressive accounting policies lead to earnings restatements. However, and more importantly, we document that any disciplinary actions against a CFO for an earnings restatement are dependent upon class-action securities litigation being brought against the firm restating its earnings. This clearly suggests that class-action securities litigation is a significant externally-imposed governance measure that explains ex post settling-up associated with the restatement event.

6. Conclusion Acknowledgements We examine the penalties imposed on CFOs as a result of earnings restatements stemming from aggressive accounting. Using a sample of 81 restatements firms and 81 control firms, we find that earnings restatements are associated with higher rates of CFO turnover and lower CFO bonus compensation. However, penalties are only imposed on CFOs when the earnings restatement led to class-action securities litigation. Further, we find that CFO penalties are contingent upon litigation even after controlling for the restatement amount scaled by assets, the litigation settlement amount, the issuance of an AAER and the market reaction to the restatement. Our results suggest that litigation is not a proxy for common measures of restatement magnitude. We also examine the periods surrounding the restatement year and find some evidence of CFO penalties in the year before the restatement year. This could be due to the firm having knowledge of the aggressive accounting problem prior to the announcement of the restatement. We find no differences between restatement firms and control firms two years before the restatement and one year after the restatement. We also examine three other groups of executives (CEOs, top-level executives and lower-level executives) and find results that are generally consistent with those for CFOs with one major exception. We find that both litigated and un-litigated restatements are associated with lower bonus payments for non-CFO executives. We fail to find evidence of higher executive turnover rates when the restatement did not result in a class-action suit.

We gratefully acknowledge the helpful comments of Larry Abbott, Davit Adut, Amy Hillman, Surya Janakiraman, Arianna Pinello, Ed Swanson, Jerold Zimmerman, participants at the 2004 KPMG Doctoral Student Accounting Association Research Meeting, participants at the 2005 AAA Annual and 2006 AAA Management Accounting Section meetings, and workshop participants at The University of Texas at San Antonio and Texas Tech University. We would also like to thank William Baber, Sok-Hyon Kang and Lihong Liang for graciously sharing their restatement data with us. A prior version of this paper was titled “Managerial Consequences of Earnings Restatements.” References Adut, D., Creedy, W., & Lopez, T. (2003). Restructuring charges and CEO compensation: A reexamination. The Accounting Review, 78(4), 169−192. Agrawal, A., Jaffe, J., & Karpoff, J. (1999). Management turnover and governance changes following the revelation of fraud. Journal of Law and Economics, 42, 309−342. Arthaud-Day, M. L., Certo, S. T., Dalton, C. M., & Dalton, D. R. (2006). A changing of the guard: Executive and director turnover following corporate financial restatements. Academy of Management Journal, 49(6), 1119−1136. Baber, W. R., Kang, S., & Liang, L. (2006). Strong boards, external governance, and accounting restatement. (April 2006; http://ssrn.com/abstract=760324) Beasley, M. S., Carcello, J. V., & Hermanson, D. R. (1999). Fraudulent financial reporting: 1987–1997: An analysis of U.S. public companies. Committee of Sponsoring Organizations of the Treadway Commission. (March 1999; http://www.coso.org/ publications/ FFR_1987_1997.PDF) Beneish, M. (1999). Incentives and penalties related to earnings overstatements that violate GAAP. The Accounting Review, 74(4), 425−457.

D. Collins et al. / Advances in Accounting, incorporating Advances in International Accounting 24 (2008) 162–171 Burks, J. J. (2008). Sarbanes–Oxley and the effect of restatements on CEO and CFO compensation and turnover. Working paper, University of Notre Dame. Carter, M., & Lynch, L. (2004). The effect of stock option repricing on employee turnover. Journal of Accounting & Economics, 37, 91−112. Carter, M., Lynch, L., & Tuna, I. (2004). The role of incentives and accounting in the design of executive compensation packages. Working paper, University of Pennsylvania Wharton School and the University of Virginia Darden School. Chen, M. (2004). Executive option repricing, incentives and retention. Journal of Finance,  59(3), 1176−1199. Collins, D., Masli, A., Reitenga, A., Sanchez, J.M. (in press). Earnings restatements, the Sarbanes–Oxley Act and the disciplining of chief financial officers. Journal of Accounting Auditing and Finance. Dechow, P., Huson, M., & Sloan, R. (1994). The effect of restructuring charges on executives' cash compensation. The Accounting Review, 69(1), 138−156. Defond, M., & Jiambalvo, J. (1991). Incidence and circumstances of accounting errors. The Accounting Review, 66, 643−655. Desai, H., Hogan, C., & Wilkins, M. (2006). The reputational penalty for aggressive accounting: Earnings restatements and managerial turnover. The Accounting Review, 81(1), 83−112. Duru, A., Iyengar, R., & Thevaranjan, A. (2002). The shielding of CEO compensation from the effects of strategic expenditures. Contemporary Accounting Research, 19, 175−193. Fee, E., & Hadlock, C. (2004). Management turnover across the corporate hierarchy. Journal of Accounting & Economics, 37, 3−38. Gaver, J., & Gaver, K. (1998). The relation between nonrecurring accounting transactions and CEO cash compensation. The Accounting Review, 73(2), 235−253. Geiger, M. A., & North, D. S. (2006). Does hiring a new CFO change things? An investigation of changes in discretionary accruals. The Accounting Review, 81(4), 781−809. Geiger, M. A., & Taylor, P. L. (2003). CEO and CFO certifications of financial information. Accounting Horizons, 17(4), 357−368. General Accounting Office (2002). Financial Statement Restatements: Trends, Market Impacts, Regulatory Responses, and Remaining Challenges. GAO-03-138. General Accounting Office (2003). Financial Statement Restatement Database. GAO-03395R. Gilson, S. (1989). Management turnover and financial distress. Journal of Financial Economics, 25, 241−262. Gilson, S. C., & Vetsuypens, M. R. (1993). CEO compensation in financially distressed firms: An empirical analysis. Journal of Finance, 48, 425−458. Hennes, K., Leone, A., & Miller, B. (2007). Management intent and CEO and CFO turnover around earnings restatements: Evidence from the post-Enron era. Working paper, Penn State University. Jones, J. M. (2005). Nurses remain atop honesty and ethics list. Gallup News Service. (December 5, 2005; http://www.galluppoll.com/content/?ci=20254&pg=1)

171

Khanna, N., & Poulsen, A. (1995). Managers of financially distressed firms: Villains or scapegoats? Journal of Finance, 50, 919−940. Kinney, W., & McDaniel, L. (1989). Characteristics of firms correcting previously reported quarterly earnings. Journal of Accounting and Economics, 11, 71−93. Krantz, M. (2002). The nine lives of ousted corporate fat cats. USAToday.com. (July 18, 2002; http://www.usatoday.com/money/covers/2002-07-18-ceos-nine-lives.htm) Lublin, J., & Forelle, C. (2004). Recovering bonuses remains infrequent despite emphasis on corporate reform. The Wall Street Journal (October 12, 2004; http://online.wsj. com/ article_print/SB109752837308242257.html) Lublin, J., & Rhoads, C. (2005). Executives on trial: The Ebbers verdict: It better not be lonely at the top in the new world of today's CEO. The Wall Street Journal (March 16, 2005; Eastern Edition, New York). McDonald, E. (2005). Wyly's war. Forbes.com. (April 25, 2005; http://www.forbes.com/ business/forbes/ 2005/0425/036_2.html) Mian, S. (2001). On the choice and replacement of chief financial officers. Journal of Financial Economics, 60(1), 143−175. Niehaus, G., & Roth, G. (1999). Insider trading, equity issues, and CEO turnover in firms subject to securities class action. Financial Management, 28(4), 52−72. Palmrose, Z., Richardson, V., & Scholz, S. (2004). Determinants of market reactions to restatement announcements. Journal of Accounting and Economics, 37, 59−89. Palmrose, Z., & Scholz, S. (2004). The circumstances and legal consequences of nonGAAP reporting: Evidence from restatements. Contemporary Accounting Research, 21(1), 139−180. Persons, O. S. (2006). The effects of fraud and lawsuit revelation on U.S. executive turnover and compensation. Journal of Business Ethics, 64(4), 405−419. Richardson, S., Tuna, I., & Wu, M. (2002). Predicting earnings management: The case of earnings restatements. (October, 2002; http://ssrn.com/abstract=338681) Sarbanes–Oxley Act of 2002. Section 304: Forfeiture of Certain Bonuses and Profits, (Public Company Accounting Reform and Investor Protection Act). Shleifer, A., & Vishny, R. W. (1997). A survey of corporate governance. The Journal of Finance, 52, 737−783. Strahan, P. E. (1998). Securities class actions, corporate governance and managerial agency problems. Working paper, Federal Reserve Bank of New York. Wahlen, J. M. (2004). Discussion of ‘The circumstances and legal consequences of nonGAAP reporting: Evidence from restatements.’ Contemporary Accounting Research, 21(1), 181−190. White, H. (1980). A heteroscedasticity-consistent covariance matrix estimator and a direct test for heteroscedasticity. Econometrica, 48, 817−838. Young, S. (2006). Court upholds Ebbers’s conviction. The Wall Street Journal, A2 (July 29, 2006; http:// online.wsj.com/article_print/SB115411140586820588.html)