The Disciplining Effect of the Internal Control Provisions of the Sarbanes Oxley Act on the Governance Structures of Firms.

Transcription

1 The Disciplining Effect of the Internal Control Provisions of the Sarbanes Oxley Act on the Governance Structures of Firms Abstract This study examines whether the internal control provisions under the Sarbanes-Oxley Act (SOX) have a disciplining effect on the governance structures of firms. We find that audit committee members and outside directors of firms that disclose material weaknesses (MWs) under Section 302 of SOX are more likely to leave the firms compared to their counterparts in a matched sample of control firms without such weaknesses, and they lose more outside directorships than their counterparts in the control firms. These results are consistent with the notion that the labor market imposes reputational penalties for internal control failures. Although the MW firms have weaker governance structures than the control firms prior to the MW detection, they show significantly greater improvement in governance structures than the control firms following the detection of these weaknesses. We also find that the market reacts positively to the improvement in audit committee size and board independence, suggesting that the improvement restores investor confidence in financial reporting. Overall, the results in this study show that the internal control provisions of SOX have a disciplining effect on the governance structures of firms. Keywords: internal controls, Sarbanes-Oxley Act, reputational penalties, corporate governance JEL codes: G30, G38, G39, M40, M42

2 The Disciplining Effect of the Internal Control Provisions of the Sarbanes Oxley Act on the Governance Structures of Firms Beng Wee Goh Singapore Management University Singapore Dan Li * TsingHua University Beijing, China lidan@sem.tsinghua.edu.cn * Corresponding author

3 1. Introduction In 2002, the U.S. Congress passed the Sarbanes-Oxley Act (SOX) to improve the quality of financial reporting and to restore investor confidence in the reliability of financial statements. The management of firms is required, by Section 302 (SOX 302) and Section 404 (SOX 404) of this Act, to assess internal controls and to report any internal control weaknesses (ICWs) that are detected. The auditor must attest to the management s assertion that the internal controls are effective and must also give an opinion on the effectiveness of the internal controls (SEC [Securities and Exchange Commission] Release Nos and ). Underlying these provisions is the belief that ICWs can result in fraudulent financial reporting, and, therefore, effective internal controls are necessary to ensure reliable financial reporting. However, companies have struggled to implement these internal control provisions and opponents of the provisions have criticized the high costs of compliance. 1 Therefore, studies that directly examine the benefits of the internal control provisions are especially useful from the perspectives of both policy makers and financial statement users. Recent research studies suggest that the internal control provisions may benefit users of financial statements by enhancing the reliability of financial reporting (Doyle et al. 2007a; Ashbaugh-Skaife et al. 2008) and by reducing the cost of equity (Ogneva et al. 2007; Beneish et al. 2008). However, no study has yet examined whether the internal control provisions have a positive impact on the governance structures of firms, the quality of which has been a major concern of regulators and investors. This study fills this gap by examining whether the internal control provisions of SOX 302 have a disciplining effect on the governance structures of firms. It specifically examines whether the directors of firms, in which ICWs are detected under SOX 302, 1 For instance, in July 2004, the Financial Executives International (FEI) surveyed 224 public firms about the direct costs of complying with Section 404 of SOX. The survey finds that the average first-year cost estimate is almost $3 million for approximately 26,000 hours of internal work and 5,000 hours of external work, plus additional audit fees of $823,200, or an increase of 53% in audit fees. 1

4 bear reputational penalties by being removed from the firms. The reputational penalties imposed on those individuals charged with monitoring internal controls can affect, ex ante, both the quality of their monitoring and their incentives to monitor (Fama and Jensen 1983), and can also shed light on the efficiency of the labor market in disciplining directors for their monitoring failures in the aftermath of SOX. Although, following the ICW detection, directors may be ousted from firms for monitoring failures, it is not clear whether they are replaced by more effective directors and, therefore, whether there is a real improvement in the governance structures of these firms. Consequently, it is important to examine not only whether directors suffer reputational penalties for monitoring failures, but also whether the new directors bring additional expertise to the firms. In order to address these questions, we also examine whether ICW firms improve their governance structures following the detection of these weaknesses. This is an important issue, given that these firms are more susceptible to fraud or financial statement errors. Also, such improvements can be seen as benefits accruing to the stakeholders of weak firms, thereby potentially justifying the high costs involved in complying with the internal control provisions. We collect a sample of 184 firms that disclosed, from August 2003 to December 2004, material weaknesses (MWs) under SOX We focus on MWs because they represent the most serious form of ICW, and, therefore, should be of greater interest to regulators and investors. Also, the disclosure of MWs is mandatory and should mitigate any self-selection 2 Auditing Standards No. 2 (PCAOB 2004) identifies three levels of internal control weakness based on the likelihood that a material misstatement of annual or interim financial statements might result. A control deficiency exists when the design or operation of a control does not allow, in the normal course of performing their assigned functions, management or employees to prevent or detect misstatements on a timely basis. A significant deficiency adversely affects the company s ability to record or report external financial data reliably in accordance with GAAP, such that there is more than a remote likelihood that a misstatement of a firm s financial statements, that is more than inconsequential, will not be prevented or detected. Material weaknesses result in more than a remote likelihood that a material misstatement of the financial statements will not be prevented or detected. 2

5 problems (Doyle et al. 2007b). 3 Furthermore, we focus on accelerated filers because of their greater market capitalization and importance. 4 Each MW firm is matched to a control firm, based on sales, industry, exchange listing, and accelerated filer status. We then examine the turnover of audit committee members and outside directors, and the loss of their outside directorships, following the MW detection. We also examine whether the governance structures of these firms improve following the MW detection. The empirical results show that the MW firms experience a greater turnover of audit committee members and outside directors than the control firms. Moreover, these results hold even after controlling for other factors, such as poor financial performance, earnings restatements, and so on. In addition, we find that the audit committee members and outside directors of the MW firms lose more outside directorships than their counterparts in the control firms. This finding suggests that the labor market infers a lower quality of monitoring associated with these individuals and, as a result, imposes further reputational penalties. Further analyses also show that the turnover of these individuals and the loss of their outside directorships increases in relation to the severity of the detected MWs. With regard to changes in governance structures, we find that, in the year prior to MW detection, the MW firms have fewer independent audit committees, lower audit committee accounting financial expertise, smaller audit committees, and fewer independent boards than the 3 Although both MWs and significant deficiencies are deficiencies in the design or operation of internal controls, significant deficiencies are less severe and are not required to be publicly disclosed under SOX 302 (SEC 2004). Hence, the disclosure of significant deficiencies is clearly voluntary. On the other hand, under SOX 302, if management identifies a MW in their controls, they are precluded from reporting that the controls are effective and must disclose the identified MW. Hence, the disclosure of MWs is effectively mandatory. According to Doyle et al. (2007b), there is some ambiguity regarding whether SOX 302 certifications require public disclosure of MWs, and whether some firms might interpret the MW disclosure requirement under SOX 302 as voluntary. The authors conclusion from the reading of the bulk of SEC guidance is that most firms are treating the disclosure as mandatory. 4 An accelerated filer (a U.S. company with market capitalization over $75 million that has filed at least one annual report with the SEC) was required to comply with the SOX 404 requirements for its first fiscal year ending on or after November 15, A nonaccelerated filer must comply with the management reporting requirements under SOX 404 for its fiscal years ending on or after December 15, 2007, and with the auditor attestation requirements under SOX 404 for its fiscal years ending on or after December 15,

6 control firms. 5 However, the MW firms improve on these governance characteristics following the MW detection, and these improvements exceed those made by the control firms. In fact, in the second year following the MW detection, there is no difference in these governance characteristics between the two groups of firms. Additional analyses show that, among the MW firms, the improvement in audit committee size and board independence is positively related to the two-year buy-and-hold abnormal returns. This finding is consistent with improvements in governance structures restoring investor confidence in the financial reporting of these firms. This study makes several contributions. First, it adds to the literature on the benefits of SOX by showing that the internal control provisions under SOX 302 have a disciplining effect on the governance structures of firms. Specifically, it shows that MW firms experience a higher turnover of directors and exhibit a significant improvement in the effectiveness of their audit committees and other board members following the MW detection. Second, this study adds to the literature on the labor market penalties imposed on directors for monitoring failures. It shows that directors suffer reputational penalties for internal control failures and that such punitive actions may provide incentives for the directors to be effective monitors ex ante. Finally, while many studies have examined the effectiveness of audit committees and boards, few have examined whether firms improve their effectiveness, especially following a monitoring failure. Farber (2005) finds that firms improve their governance structures after a fraud event in order to restore public trust. This study adds to the literature in this area by showing that firms improve their governance structures after internal control failures. 5 Accounting financial expertise refers to experience as a public accountant, auditor, principal or chief financial officer, controller, or chief accounting officer. Nonaccounting financial expertise refers to the experience as a CEO, president, general partner, or managing director of a for-profit corporation. The definition of financial expertise in the earlier version of SOX included only the formal definition of expertise, but the final version of SOX was expanded to include both definitions of expertise. A detailed discussion on the financial expertise of the audit committee can be found in Section IV of this study. 4

7 The rest of this study is organized as follows. Section 2 reviews related studies and develops the hypotheses. Section 3 describes the sample collection procedures and composition. Section 4 presents the research design and empirical results, and Section 5 presents additional analyses. Section 6 concludes and discusses limitations of this study. 2. Related studies and hypothesis development There has been an intense debate regarding the costs and benefits of the internal control provisions under SOX 302 and SOX 404. Critics of the provisions have maintained that the costs of compliance are too high to justify their benefits. Although the debate has generated much research interest in these internal control provisions, studies that directly examine their benefits are limited. Early studies find that firms with ICWs tend to be smaller, less profitable, more complex, growing rapidly, undergoing organizational changes, and so on (Ashbaugh-Skaife et al. 2007; Ge and McVay 2005; Doyle et al. 2007b). However, these studies do not examine the costs or the benefits of the internal control provisions. The studies that examine the relation between internal controls and a firm s earnings quality generally document that firms with ICWs tend to have lower accruals quality, as measured by the extent to which accruals are realized as cash flows (Doyle et al. 2007a; Ashbaugh-Skaife et al. 2008). Ashbaugh-Skaife et al. (2008) further find that the remediation of ICWs improves accruals quality. Finally, recent studies show that firms disclosing ICWs have a marginally higher cost of equity than firms disclosing no such weaknesses (Ogneva et al. 2007; Beneish et al. 2008). Overall, these later studies suggest that the internal control provisions benefit financial statement users by improving the quality of financial reporting and by reducing the cost of capital of firms. In contrast to the above studies, this study investigates whether the implementation of the internal control provisions under SOX 302, namely the requirement to disclose known MWs in internal control, has a positive impact on the governance structures of firms. The study 5

8 specifically examines whether the directors of firms which disclose ICWs bear reputational penalties for the breakdowns in internal control by being removed from the board. It also examines whether, following the ICW detection, these firms make substantial improvements in their governance structures, in terms of the effectiveness of their audit committees and boards. Empirical evidence of reputational penalties for internal control failures and of improvements in governance structures subsequent to the ICW disclosure can be seen as direct benefits accruing from the internal control provisions under SOX 302. It may be costly for firms to punish and remove incumbent directors following the detection of ICWs; for example, they may have to expend time and resources looking for possible replacements for the incumbent directors. Furthermore, firms may want to maintain continuity of directorship in order to remediate the weaknesses detected, as new directors may be unable to effect a prompt remediation if they are not familiar with the internal control systems. However, it seems reasonable to believe that, when ICWs are detected, some form of punitive action would be imposed on the individuals charged with monitoring internal controls. Due to the separation of ownership from management within the large corporation (Fama and Jensen 1983), the board of directors plays an important role in monitoring internal controls to protect the shareholders interests from the management s opportunism. Given that a breakdown in internal controls indicates an oversight failure on the part of the board directors, labor market forces could impose disciplining actions on these directors following ICW detection. The existing evidence that is consistent with such reputational penalties being associated with monitoring failures includes that of Gilson (1990), who finds that disciplinary actions are taken against directors in firms that either file for bankruptcy or privately restructure their debt, and Srinivasan (2005), who finds, in firms that overstate earnings, a 48.1 percent turnover of outside directors within three years of the restatement announcement. 6

9 The audit committee of a firm may be especially culpable for ICWs, as it plays a vital role in monitoring internal controls. According to Reinstein et al. (1984), outside non-auditcommittee directors are likely to demonstrate the fulfillment of their fiduciary duties by stating that they relied upon the representations of the audit committee on issues regarding the adequacy of financial reporting. Consequently, the non-audit-committee directors effectively shift some of the risk of potential financial misstatements to the audit committee (Abbott et al. 2003), and the audit committee would seek to mitigate risk by diligently performing the oversight of financial reporting. One way in which the audit committee ensures adequate financial reporting is by monitoring the internal controls. Furthermore, under Section 301 of SOX, audit committees of public companies are required to establish procedures for the receipt, retention, and treatment of complaints received regarding accounting, internal controls, and auditing. Hence, audit committee members are likely to take the blame for internal control failures and to suffer some form of punitive actions. Recent concerns about weak internal controls leading to accounting improprieties and the publicity surrounding several high-profile cases, such as Enron and Worldcom, have heightened the attention of investors and regulators to internal controls. This increased scrutiny by investors and regulators is likely to intensify further the pressures on firms to punish directors for internal control failures. Although ICWs are not outright fraud, weak internal controls and fraud are closely related. For instance, Bell and Carcello (2000) examine the effectiveness of the riskfactor examples presented in SAS (Statement on Auditing Standards) No. 53 in distinguishing between fraud and nonfraud engagements, and find that a weak internal control environment is positively related to fraud. In 1999, a study commissioned by the Committee of Sponsoring Organizations of the Treadway Commission found that a poor internal control environment contributed to the occurrences of fraud that were documented during the period 1987 to

10 Given that ICWs may be seen as a precursor of fraud, we expect that internal control failures will exert pressure on firms to remove ineffective directors. Finally, Beneish et al. (2008) find that ICWs disclosed under SOX 302 are associated with negative announcement abnormal returns of -1.8 percent and that ICW firms experience an abnormal increase in equity cost of capital of 68 basis points. Hammersley et al. (2008) also find some evidence that the market reacts negatively to firm announcements of SOX 302 ICW disclosures. These negative stock market reactions could provide further impetus for ICW firms to punish their audit committees and board members. Compared to firms without ICWs, we expect these individuals to be more likely to leave the ICW firms if punitive actions are taken following the detection of weaknesses. We focus only on outside directors, rather than on the full board, because the academic literature and the regulators emphasize the role of these directors in monitoring financial reporting. Furthermore, Fama and Jensen (1983) contend that the risks and rewards of board memberships apply to these directors. Hence, throughout this paper, the term outside directors refers to the directors that are nonemployees. If the external labor market infers the lower quality of monitoring associated with the audit committee members and outside directors of ICW firms, we expect these individuals to lose more outside directorships than their counterparts in other firms without such weaknesses. 6 For instance, Srinivasan (2005) finds that outside directors of firms that overstate earnings lose 25 percent of their positions on other boards, and Gilson (1990) finds that the directors who resign from firms that either file for bankruptcy or privately restructure their debts hold 6 Fama and Jensen (1983) contend that board membership confers benefits to directors and that outside directors are rewarded by the reputation they develop as expert monitors. Specifically, outside directors use their directorships to signal to internal and external markets that (1) they are decision experts, (2) they understand the importance of diffuse and separate decision control, and (3) they can work with such decision control systems. Good performance by board directors has the potential to lead to better opportunities for the directors going forward, such as gaining more board seats and external job opportunities. On the other hand, Fama and Jensen argue that, when monitoring fails, labor market forces impose disciplinary actions on directors. Such reputational penalties include not only the loss of their board positions, but also extend to the loss of directorships in other firms. 8

11 significantly fewer seats on other boards following their departure. Based on the above discussions, we make the following hypotheses: Hypothesis 1a: Firms with ICWs experience a greater turnover of audit committee members and outside directors following the detection of ICWs, compared to firms without such weaknesses. Hypothesis 1b: Audit committee members and outside directors in firms with ICWs lose more outside directorships in other public companies following the detection of ICWs, compared to their counterparts in firms without such weaknesses. Although, following the ICW detection, directors may be ousted from firms for monitoring failures, it is not clear whether they are replaced by more effective directors, and, therefore, whether there is any real improvement in the governance structures of these firms. Following the detection of ICWs, a firm is likely to have sufficient incentives to improve its governance structures. As one board function is to monitor the performance of top managers (Fama and Jensen 1983), the existence of ICWs may reveal a failure by the board to monitor these managers effectively. The detection of ICWs also suggests a failure by both the board and the audit committee to ensure effective internal controls and adequate financial reporting. Given this, an improvement in the effectiveness of either the board or the audit committee, following the detection of ICWs, can increase the net returns to managerial oversight, increasing the value of the services of new board members. Following the detection of ICWs, the ICW firm is also likely to lose its reputation because the existence of these weaknesses indicates that the firm did not uphold its part of an implicit contract with the stakeholders. The negative publicity that is likely to follow the announcement of internal control failures prompts the firm to find ways to recover lost reputation capital (Agrawal et al. 1999). New board members can bring reputational capital with them that has a relatively high value, especially to the firm s investors or suppliers. The improvement in the effectiveness of the board and the audit committee can be a cost-effective way for the firm to 9

12 reinvest in, and to reestablish, the firm s reputation capital. For instance, Farber (2005) finds that firms, in which fraud has been detected, improve their governance structures subsequent to the fraud detection in order to restore trust. As previously mentioned, studies have shown that, when ICWs are disclosed, firms experience a decline in their stock prices (Beniesh et al. 2008; Hammersley et al. 2008). Therefore, an improvement in governance structures may also help the firm to recover its performance (Agrawal et al. 1999). Rosenstein and Wyatt (1990) document positive mean abnormal returns surrounding the announcement dates of outside director appointments, and Farber (2005) finds a significant positive relation between an increase in board independence and long-run buy-and-hold abnormal returns. To the extent that the improvement in governance structures helps to signal a firm s commitment to restoring effective internal controls and improving financial reporting quality, we expect firms to have strong incentives to make such improvements following the detection of ICWs. Lastly, improving the governance structure also helps to limit the firm s legal liability resulting from either a failure to ensure effective internal controls or a sharp stock price decline triggered by such internal control failures. It is likely that the recent high-profile corporate failures and the focus on internal control breakdowns has heightened public awareness of the governance structures of firms. As a result, firms across the board are likely to improve their governance structures, and, therefore, it is necessary to compare the improvement in the governance structures of ICW firms to those in a sample of control firms. If the ICW firms have greater incentives than the control firms to improve their governance structures, then we can expect them to show an improvement in their governance structures relative to the control firms. Hence, the following is hypothesized: Hypothesis 2: Corporate governance structures, that is the effectiveness of the audit committee and the board of directors of ICW firms, improve following the detection of ICWs, relative to firms without such weaknesses. 10

13 3. Sample 3.1 Selection of sample and control firms We identify sample firms that disclose MWs under SOX 302 from Doyle et al. (2007b), Compliance Week, and AuditAnalytics. 7 We focus on SOX 302 weaknesses because they are disclosed earlier, and, therefore, any corporate governance changes observed are more likely to be attributable to SOX 302, rather than to SOX 404, weaknesses. As mentioned in the introduction, we focus on firms that disclose MWs because the reporting of MWs is mandatory, whereas the reporting of significant and control deficiencies is not. Moreover, using firms that disclose significant and control deficiencies may create self-selection problems. Finally, we choose firms that are accelerated filers, as these firms have greater market capitalization and, hence, have a greater economic impact and arouse more public interest when MWs are detected. Panel A of Table 1 summarizes the sample collection procedure. The initial sample comprises 502 firms that disclosed at least one MW under SOX 302 during the period August 2003 to December Then, we exclude the following firms from this sample: 152 nonaccelerated filers, 72 firms that subsequently terminated their securities registration, 26 firms that delayed filing their 10-Ks, 21 firms without the proxy statements to obtain the governance data, 13 firms with MWs dating back to before the 2002 fiscal year,12 firms that are foreign issuers, 9 firms without the second SOX 404 reports, and 8 firms that are subsidiaries of other firms within the sample. This procedure yields a sample of 189 firms that report at least one MW from August 2003 to December We thank Jeffrey Doyle, Weili Ge, and Sarah McVay for sharing the data. The data can be found at Compliance Week ( is an internet newsletter on corporate governance, risk, and compliance that publishes monthly reports on firms reporting internal control weaknesses since November AuditAnalytics ( is an online market intelligence service from Ives Group Inc, which keeps track of all firms disclosing internal control problems after SOX came into effect. 11

14 The sample period is subsequent to the passage of SOX. As directors may voluntarily opt out of directorships (possibly due to greater responsibilities or to litigation risks) or firms may voluntarily improve their governance structures, it is important to eliminate these alternative explanations by comparing the turnover of directors and the improvement in governance structures in the selected MW firms against a set of control firms without these weaknesses. Each of the 189 MW firms is matched to a firm that does not report any ICWs during the period November 2002 to December 2006 and which is also an accelerated filer from the same fourdigit industry SIC (Standard Industrial Classification) code and the same stock exchange, and with net sales within + or 25 percent of the MW firm s net sales. If we cannot find a match, we relax the criteria to either a three-, two-, or one-digit SIC code. Five sample firms are excluded because we cannot find a control firm, even in the one-digit SIC code group. Therefore, the final usable sample consists of 184 firms. These firms are matched with the SIC codes as follows: four-digit (102 firms), three-digit (15 firms), two-digit (36 firms), and one-digit (31 firms). Panel B of Table 1 shows the composition of the MW and the control firms based on the industry SIC codes. This panel shows a relatively high concentration of MW firms in the manufacturing and services industries. This result is consistent with Krishnan (2005) who finds a relatively high number of firms with internal control problems in these industries. Panel C of Table 1 shows the composition of the MW firms based on their exchange listing. This panel shows that the sample firms are mostly listed on the larger stock exchanges; this is most likely due to the accelerated-filer status of both the MW and the control firms. Table 2 summarizes the definitions of the variables used in this study. 3.2 Summary statistics Panel A of Table 3 reports various summary statistics for the MW and the control firms. The financial and governance data is measured for the year prior to the MW detection. The 12

15 financial data show that both groups of firms do not differ significantly in their mean net sales and their total assets (LGTA), indicating that the matching based on size is successful. In terms of financial performance, the mean buy-and-hold returns from months -12 to +12 relative to the MW detection (STKPERF) are 0.56 and 0.62 for the MW and the control firms, respectively, and this difference is not statistically significant. However, the MW firms are in higher financial distress, as indicated by the ZFC measure (p < 0.05, two-tailed). Also, the MW firms are more likely than the control firms to undergo either mergers, acquisitions, or restructuring (MARESTR), have more operating segments (SEGMENTS), and have foreign operations (FOREIGN) than the control firms (p < 0.05, two-tailed). Turning to the governance data, Panel A in Table 3 shows that there is no significant difference between the two groups of firms in terms of (1) the mean age and tenure of their audit committee members (ACAGE; ACTENURE) and outside directors (BDAGE; BDTENURE), and (2) the proportions of split CEO-Chairman leadership structure (DUALITY), blockholder ownership (BLOCKOWN), inside ownership (INSIDEOWN), and institutional ownership (INSTOWN). However, 55 percent and 14 percent, respectively, of the MW and the control firms have at least one earnings restatement within six months of the detection of MW, and this difference is statistically significant (p < 0.01, two-tailed). 4. Research design and empirical results 4.1 Turnover of audit committee members and outside directors (Hypothesis 1a) Panel B of Table 3 provides the descriptive statistics and univariate test results on the turnover of audit committee members and outside directors. This panel shows that the proportion of audit committee members and outside directors who leave a firm within two years of the MW detection is significantly higher in the MW firms than in the control firms (p < 0.01, one-tailed). In the MW firms, 35 and 33 percent, respectively, of audit committee members and outside 13

16 directors leave the firm within two years of the MW detection; the corresponding percentages in the control firms are 19 and 21. ACTURNOVER (BDTURNOVER) is coded 1 if more than half of the audit committee members (outside directors) in the year before the MW detection leave the firm within two years of the MW detection, and 0 otherwise. The results using the indicator variables are consistent with those using proportions. Specifically, 27 percent (24 percent) of the MW firms and 4 percent (5 percent) of the control firms have more than half of the audit committee members (outside directors) leave the firm within two years of the MW detection, and the differences are statistically significant (p < 0.01, one-tailed). Pr (ACTURNOVER/ BDTURNOVER=1) = a + b 1 MWF + b 2 STKPERF + b 3 ZFC + b 4 RESTATE + b 5 LGTA + b 6 MARESTR + b 7 ACAGE/ BDAGE + b 8 ACTENURE/ BDTENURE + b 9 DUALITY + b 10 BLOCKOWN + b 11 INSIDEOWN + b 12 INSTOWN + ε We use the above logistic regression to test whether the MW firms experience a greater turnover of their audit committee members and outside directors than the control firms. MWF is coded 1 for MW firms, and 0 for control firms. If the MW firms experience a greater turnover of audit committee members and outside directors subsequent to the MW detection, then the coefficient b 1 is expected to be positive and significant. Gilson (1990) finds that firms in greater financial distress have a higher turnover of board directors. Hence, we control for firm performance using the variables STKPERF (the raw buy-and-hold returns from months -12 to +12 relative to the MW detection) and ZFC (the financial distress score calculated from the probit coefficients of Zmijewski [1984]). As earnings restatement firms are more likely to experience turnovers of audit committee members and outside directors (Srinivasan 2005), we include the indicator variable RESTATE, which is coded 1 if a firm announces one or more earnings restatements within six months of the MW detection, and 0 otherwise. As the matching based on size is not perfect, we also control for firm size using the variable LGTA, which is the natural log of the firm s total assets. Likewise, as firms undergoing 14

17 mergers, acquisitions, or restructuring can experience significant board changes, we include MARESTR, which is an indicator variable coded 1 if the firm reports any restructuring charges or if the AFTNT1 file in Compustat indicates the presence of a merger and acquisition, and 0 otherwise. We control for the characteristics of individuals using ACAGE (BDAGE) and ACTENURE (BDTENURE), which are, respectively, the mean age and tenure of the audit committee members (outside directors). In addition, we control for CEO duality, as CEOs who are also chairmen of the board have a greater influence on the board and may affect turnover (Beasley 1996; Dechow et al. 1996); the variable DUALITY is coded 1 if the CEO and Chairman positions are held by different individuals, and 0 otherwise. As higher inside ownership may also increase board entrenchment (Desai et al. 2006), we control for inside ownership using the variable INSIDEOWN, which is the total percentage of stock held by the management and directors. Lastly, firms with higher levels of blockholder ownership and institutional ownership are more likely to be better monitored (Shleifer and Vishny 1996) and, therefore, to remove ineffective directors from the firm. Therefore, we include the variables BLOCKOWN (the percentage of stock held by blockholders of the firm) and INSTOWN (the percentage of stock held by all institutional owners of the firm). 8 Table 4 shows the regression results on the relation between the incidence of MW and the turnover of audit committee members. 9 The descriptive statistics in Table 3 show that 55 percent of the MW firms also report earnings restatements, and Srinivasan (2005) finds that firms that restate earnings experience a higher turnover of audit committee members and outside directors. Hence, in Model 1, we first present the results without the indicator variable MWF, in order to 8 We hand collect the information on the stock ownership of the blockholders, management, and directors from the proxy statements. Data on institutional holdings are obtained from the CDA Investment Technologies Spectrum database, which is derived from the SEC Form 13-F disclosure forms reported quarterly to the SEC. 9 The Pearson correlations coefficients among the independent variables range from (between INSIDEOWN and LGTA) to 0.64 (between BLOCKOWN and INSIDEOWN). The VIFs of all the independent variables are under 2.5, suggesting that multicollinearity is not a concern. 15

18 examine whether restating firms are more likely to experience a higher turnover of their audit committee members. In Model 2, MWF is included to examine whether, after controlling for earnings restatement, the existence of MW results in a higher turnover of audit committee members. Finally, Model 3 presents the results by excluding the restating firms. Consistent with the findings of Srinivasan (2005), the variable RESTATE is positive and significant (p-value <0.01, one-tailed) in Model 1, suggesting that earnings restating firms are more likely to experience a higher turnover of audit committee members. However, after including MWF in Model 2, RESTATE becomes insignificant and MWF is positive and significant (p < 0.01, one-tailed). This result suggests that MW firms are more likely to experience a significant turnover of their audit committee members following the MW detection. Focusing on the sample of non-earnings-restating firms in Model 3, MWF is still positive and significant (p < 0.01, one-tailed), which shows that the higher turnover of audit committee members is driven by the existence of MW and not by earnings restatement. The variable STKPERF is negative and significant (p < 0.05, one-tailed) in all models and ZFC is positive and significant (p < 0.10, one-tailed) in Model 1; these results suggest that poor financial health increases the likelihood of audit committee members leaving the firm. Lastly, DUALITY is positive and significant (p < 0.05 in Models 1 and 2, one-tailed), suggesting that audit committee members are more likely to leave the firm when the positions of CEO and Chairman are split. Table 5 repeats the above analyses by looking at the outside directors of the board to examine whether the reputational penalties extend to the entire board. The results are similar to those of Table 4. Specifically, RESTATE is positive and significant (p-value < 0.05, one-tailed) in Model 1. When MWF is included in Model 2, RESTATE becomes insignificant and MWF is positive and significant (p-value < 0.01, one-tailed). In Model 3, MWF is again positive and significant (p-value < 0.01, one-tailed). Taken together, these results suggest that the existence of 16

19 MW results in a higher turnover of outside directors, even after controlling for earnings restatement, and that the reputational penalties resulting from MWs extend from the audit committee members per se to the outside directors of the entire board. Also, the results in Table 5, like those in Table 4, show that outside directors are more likely to leave the firm when the financial health of the company is poorer and the positions of CEO and Chairman are held by different individuals. In sum, the empirical results above show that audit committee members and outside directors bear reputational penalties for internal control failures. These individuals are more likely to leave the firm following the MW detection. If internal control failures reflect the lower quality performance of these individuals in monitoring and the external labor market can infer this, such reputational penalties can extend into multiple arenas. The next section presents the results relating to whether audit committee members and outside directors in the MW firms suffer greater losses in outside directorships than their counterparts in the control firms. 4.2 Loss of outside directorships (Hypothesis 1b) Panel B of Table 3 presents the descriptive statistics on the mean number of outside directorships held by the audit committee members and outside directors in both the MW and the control firms. 10 In the year prior to the MW detection, the outside directors of the MW firms hold significantly more outside directorships (BDSEATBEF) than their counterparts in the control firms (p < 0.05, two-tailed). This greater number of outside directorships may have reduced the 10 Information on the outside directorships in public companies is obtained from the proxy statements. For directors who leave the firm, we use AuditAnalytics to download the director information of all SEC registrants, as of December 15, This allows me to track any public company boards that these directors sit on in the two years following the MW detection. Then, we refer to the proxy statements of these public companies to obtain the directorship information. As an additional check, we also use to track any public companies on whose boards these directors sit during the two years following the MW detection. ZoomInfo is a free internet summarization search engine that provides comprehensive information on over 31 million business professionals and two million companies. The website tracks information from millions of online sources such as Web sites, press releases, electronic news services, and SEC filings and summarizes the information into a comprehensive format. 17

20 internal control monitoring effectiveness of the outside directors in the MW firms. During the period from the year before to the second year after the MW detection, the audit committee members of the MW firms and the control firms lose, respectively, an average of 0.19 and 0.02 of their outside directorships (ACSEATLOSS). During the same period, the outside directors of the MW and the control firms lose, respectively, an average of 0.21 and 0.05 of their outside directorships (BDSEATLOSS). The differences in ACSEATLOSS and BDSEATLOSS between the two groups of firms are both significant (p < 0.01, one-tailed). Finally, in the second year following the MW detection, the mean number of outside directorships held by audit committee members (ACSEATAFT) and outside directors (BDSEATAFT) in the MW and the control firms are not significantly different. ACSEATLOSS/ BDSEATLOSS = a + b 1 MWF + b 2 STKPERF + b 3 ZFC + b 4 RESTATE + b 5 LGTA + b 6 ACSEATBEF/ BDSEATBEF + b 7 ACAGE/ BDAGE + b 8 ACTENURE/ BDTENURE + ε We use the above OLS regression model to test the relation between the incidence of MWs and the loss of outside directorships by the audit committee members and outside directors. Hypothesis 1b predicts the coefficient b 1 to be positive and significant. We control for the effect of poor financial performance on the loss of outside directorships using STKPERF and ZFC. We include RESTATE because the audit committee members and outside directors of restating companies suffer greater losses in outside directorships (Srinivasan 2005). We also control for size using the log of the total assets of the firm (LGTA). As directors with more outside directorships are likely to lose more directorships, we control for ACSEATBEF and BDSEATBEF. Lastly, we control for the mean age and tenure of the audit committee members (ACAGE; ACTENURE) and outside directors (BDAGE; BDTENURE). Models 1 and 2 of Table 6 show, respectively, the regression results for the loss of outside directorships of the audit committee members and outside directors. Consistent with the 18

21 univariate test results, MWF is positive and significant (p < 0.01, one-tailed), suggesting that the incidence of MWs increases the number of outside directorships lost by both the audit committee members and outside directors. Both ACSEATBEF and BDSEATBEF are positive and significant (p < 0.01, one-tailed), which is consistent with the loss of outside directorships increasing with the number of outside directorships held. Contrary to expectations, ACTENURE and BDTENURE are positive and significant (p < 0.05, one-tailed), suggesting that longer tenures on the board increase the loss of outside directorships. One possible explanation for this finding is that longer tenures increase the bond and commitment of these individuals to the firm, which, in turn, increase the likelihood of them losing or giving up directorships in other firms. Overall, the results in this section and the previous section are consistent with the argument that the labor market imposes reputational penalties on audit committee members and outside directors for internal control failures. The reputational penalties for internal control failures extend from the loss of positions within the MW firms to the loss of outside directorships in other public companies. In the next section, we present the empirical results on whether, following the detection of MWs, the MW firms improve the effectiveness of their governance structures, that is the effectiveness of their audit committees and boards of directors. 4.3 Improvement in Governance Structures (Hypothesis 2) Determinants of the effectiveness of corporate governance structures First, we establish the factors that determine the effectiveness of both the audit committee and the board of directors. Research has shown that audit committees that are more independent reduce the likelihood of financial reporting misstatements (Abbott et al. 2000) and increase the independence of the external auditor (Carcello and Neal 2000, 2003). Therefore, we measure audit committee independence using ACINDP, an indicator variable coded 1 if the audit committee comprises fully independent directors, and 0 otherwise. Higher financial expertise of 19

22 the audit committee is found to be associated with fewer financial reporting problems (McMullen and Raghunadan 1996; Abbott et al. 2002). However, one controversy over SOX is the definition of financial expertise used to define the expertise of an audit committee member. 11 Given this controversy, we examine whether there is an improvement in the financial expertise of the audit committee using two definitions, namely accounting and nonaccounting financial expertise. The variables ACCEXP and NONACCEXP represent, respectively, the proportion of audit committee members with accounting financial expertise (i.e., prior accounting-related experience with SEC financial reporting, for example as a public accountant, auditor, principal financial or accounting officer, or controller) and with nonaccounting financial expertise (i.e., expertise gained through experience supervising employees with financial reporting responsibilities and overseeing the performance of companies, for example as a company president, CEO, general partner, managing director, or general principal). Larger audit committees are more likely to generate substantial discussions and to consider emerging issues (DeZoort et al. 2002), and this results in higher-quality financial reporting (Felo et al. 2003) and lower bond yield spreads (Anderson et al. 2004). we measure audit committee size using ACSIZE, which represents the number of audit committee members. Also, greater audit committee diligence is found to be associated with lower incidences of fraud (Beasley et al. 2000) and with the employment of an industry specialist auditor (Abbott and Parker 2000). Hence, we capture the audit committee s diligence using ACMEET, which represents the number of times the audit committee meets in a fiscal year. 11 Although SOX requires that firms disclose whether they have a financial expert on the audit committee, controversy arose over the appropriate definition of financial expertise. The initial SOX promulgations recommended a fairly narrow definition of financial expertise that focuses on whether the director has prior accounting related experience with SEC financial reporting and suggests that such directors will have work experience as a public accountant, auditor, principal financial or accounting officer, or controller ( accounting financial expertise ). The final version of the SOX provisions effectively expands the definition of financial expertise by also including the expertise gained through experience supervising employees with financial reporting responsibilities and overseeing the performance of companies ( nonaccounting financial expertise ). This wider definition hence captures directors who have prior experience as company presidents and CEOs. 20

23 Boards that are more independent can reduce the likelihood of fraud (Beasley 1996) and earnings management (Klein 2002), and we therefore define BDINDP as the proportion of board members who are independent. Smaller boards are more effective monitors because they have a controlling function and reduce coordination and processing problems (Jensen 1993). Hence, we measure board size using BDSIZE, which represents the number of directors on the board. Previous studies have shown that greater diligence by the board increases firm performance (Vafeas 1999; Klein 1998); the variable BDMEET, which represents the number of times the board meets in a fiscal year, measures board diligence. Finally, separating the positions of CEO and Chairman of the board may strengthen the effectiveness of the board (Jensen 1993) and reduce the likelihood of SEC enforcement actions (Dechow et al. 1996). We capture the duality of the CEO and Chairman positions using the indicator variable DUALITY, which is coded 1 if the two positions are held by different individuals, and 0 otherwise. All of the governance information is obtained from the firms proxy statements Univariate test and regression results Panels A and B of Table 7 present, respectively, the univariate comparisons of the changes in the characteristics of the audit committee and the overall board, from the year before to the second year after the MW detection. Columns 1 to 4 show the governance characteristics in the year relative to the year of MW detection (year t), as well as the t-statistics to compare the differences in governance characteristics between the MW and the control firms. Columns 5, 6, and 7 show, respectively, the changes in each governance characteristic from year t-1 to years t, t+1, and t+2, as well as the t-statistics to compare the differences in the changes in each governance characteristics between the MW and the control firms. Column 8 shows the t- statistics, which test whether the changes in the governance characteristics from year t-1 to year t+2 are significant. 21