The Relation between Earning Informativeness, Earnings Management and Corporate Governance in the Pre- and Post-SOX periods

Transcription

1 The Relation between Earning Informativeness, Earnings Management and Corporate Governance in the Pre- and Post-SOX periods Jui-Chin Chang* Department of Accounting School of Business Howard Universy th Street NW Washington DC Tel: (202) jchang@howard.edu Huey-Lian Sun Department of Accounting & Finance School of Business & Management Morgan State Universy 1700 Cold Spring Lane Baltimore, MD Tel: (443) HueyLian.Sun@morgan.edu Working Paper November 2008 *Correspondent author Electronic copy available at:

2 The Relation between Earning Informativeness, Earnings Management and Corporate Governance in the Pre- and Post-SOX Periods Abstract We pos that the SOX regulations on aud commtee independence and other corporate governance have improved the qualy of accounting earnings. Using sample firms earnings informativeness and earnings management to measure the qualy of accounting earnings, we find significantly posive (negative) relations between earnings informativeness (earnings management), aud commtee independence and financial experts on aud commtee in the post-sox period while we find no significant relations in the pre-sox period. Also, a fully independent aud commtee and a majory independent board are found to complement each other in increasing (decreasing) earnings informativeness (earnings management) in the post-sox period. Furthermore, by comparing the pre- and post-sox results, we report a magnude change of the relation between earnings management (earnings informativeness) and Big 4 audors (dualy of CEO, Big 4 audors and overall corporate governance). Overall, our findings indicate that the market perception of corporate governance mechanisms in monoring the qualy of accounting earnings has changed after SOX. In addion, we find that the effectiveness of corporate governance in monoring managerial behaviors on earnings management has also improved after SOX. Key words: Corporate Governance, SOX, Earnings Informativeness, Earnings Management Electronic copy available at:

3 The Relation between Earning Informativeness, Earnings Management and Corporate Governance in the Pre- and Post-SOX Periods 1. Introduction Effective corporate governance to oversee financial reporting process is fundamental to preserve investors confidence in the capal markets. Prior to the Sarbanes-Oxley Act (hereafter, SOX or the Act), even though the national stock exchanges and the national securies market associations (SROs) have required publicly traded firms to have a fully independent aud commtee wh at least three members under the Blue Ribbon Commission s (BRC) recommendations, the Securies and Exchange Commission (SEC) gave firms discretion to appoint non-independent directors to their aud commtees. Meanwhile, small firms were exempted from the requirements of three independent members and were allowed to have only two members in their aud commtee. 1 Therefore, prior to SOX was up to firms to voluntarily set up independent board, independent aud commtee or independent audors in their corporate governance structure (Michell, 2003). The Act was implemented in the midst of the various financial scandals that occurred in the early 2000s, when investors became knowledgeable of the questionable tactics used by firms to manipulate financial reporting. In response to the financial scandals, the Act aims to restore investors confidence in the capal markets. Consequently, there are changes in regulations which affect companies leadership and corporate governance structures as well. Under the Act, the corporate board is expected to play an essential role in providing effective oversight over a company s financial reporting system. First, the SEC mandates all issuers in the SROs to comply wh the 1 Refer to the SEC, Final Rule: Aud Commtee Disclosure, 17CFR parts 210,228, 229 and 240; Release No The definion of small firms was firms whose total assets were $5 million or less. 1 Electronic copy available at:

4 requirement of a fully independent aud commtee. Firms are no longer allowed to appoint non-independent directors to their aud commtee. The deadline to comply wh the new listing rules is by January 15 or October 31, 2004, whichever is applicable. 2 In implementing the aud commtee independence requirements, the SROs may delist any issuer that does not comply wh the new listing rules. Furthermore, in order to improve aud commtee effectiveness in monoring the credibily of accounting reporting, the SEC also requires firms to disclose financial experts on their aud commtees. 3 In addion, criques on financial scandals address the need for managers and audors to be held responsible for the credibily of financial reporting. Some cricize that the outside directors are ignorant of what is inside of the company. There are also concerns that the CEOs serving as the chair of their board may have final words to influence the financial reporting decisions. Due to these criques, the Act requires CEOs of all companies to certify their financial statements. After the fall of Arthur Anderson, the Act further restricts non-aud services to ensure audors independence in auding financial statements. 4 As noted by recent studies (e.g., Cohen, Dey and Lys, 2007; Zhang, 2007; Romano, 2005), firms now need more time and incur higher costs to fulfill the strict requirements under SOX, and the compliant costs may outweigh the compliant benefs. Other studies (La Porta et al., 2000; Coffee Jr., 2002; Li, Pincus and Rego, 2008), however, argue that government-enforced regulations are necessary in providing 2 Refer to footnote 2. Under the SEC s new definion, small firms are firms whose revenues are less than 25 millions. 3 Refer to the SEC, Disclosure required by Section 406 and 407 of the Sarbanes-Oxley Act of 2002;17 CFR Parts 228, 229, and 249; Release Nos ; ; File No. S Meanwhile, smaller firms are required to comply after December 15, Refer to Section 201, Services outside of the scope of practice of audors, of the Sarbanes-Oxley Act of

5 transparent and truthful disclosures when there is no other effective and economic way to enforce private contracts through the court system. Yet, the issue whether the SOX regulations have posively impact on firms disclosure of their corporate governance in monoring the qualy of financial reporting is highly controversial, and remains an empirical question. The purpose of this study is to examine the effect of mandatory disclosure of aud commtee independence and other corporate governance information on the qualy of firms reported accounting earnings after SOX. We use earnings informativeness and earnings management as two proxies for the qualy of accounting earnings. Earnings informativeness is measured by the marginal effect of the market reaction to firms earnings announcements condioned on the unexpected earnings and corporate governance variables. We measure earnings management by discretionary accruals. We also compare the magnude changes of relations between earnings informativeness (earnings management) and corporate governance variables in the pre- and post-sox periods in examining the effect of SOX regulations on the firms qualy of accounting earnings. Our findings show significantly that earnings informativeness is posively related to aud commtee independence, financial experts on aud commtees, separation of CEO and chair of board, and a firm s overall corporate governance in the post-sox while their relations were not significant in the pre-sox period. The magnude change in the relations suggests that after SOX, the earnings informativenss increases due to the change of the market perception of these corporate governance mechanisms on the qualy of firms accounting earnings. We also find significantly negative relationship 3

6 between earnings management and aud commtee independence, and financial experts on aud commtees after the passage of SOX but we do not find significant relations in the pre-sox period. The change of the relations implies that the above corporate governance mechanisms, not effective in monoring managerial behavior on earnings management before SOX, become effective after SOX. In addion, we find that the relation between the earnings management and Big 4 audors changes from a significantly negative relation in the pre-sox period to an insignificant relation in the post-sox period, suggesting that Big 4 audors might have shifted their oversight duties of firms earnings management to the firms aud commtee after SOX. This study contributes to the existing corporate governance lerature from several perspectives. First, this study is among the first to examine the effect of the Act on earnings informativeness in the post-sox period. Second, we are also among the first to examine the relations between earnings management and the effectiveness of aud commtee independence and the other corporate governance functions in the post-sox period. Third, compared wh previous studies using small sample size and single indicator for corporate governance, we use a relatively larger sample size to generalize the relations between earnings informativeness (earnings management), and aud commtee independence and other corporate governance functions. In addion, since firms might set up their corporate governance functions from different dimensions to form an optimal monoring structure, we construct an aggregated corporate governance score to examine s relation wh earnings informativeness and earnings management. Last, we compare the pre-sox and post-sox relationship between earnings informativeness (earnings management) and corporate governance functions for the 4

7 effect of the SOX regulations on the qualy of accounting earnings. Our findings provide insights on how regulations impact the market perception of corporate governance and further evidence on the corporate governance in deterring earning management. The remaining sections proceed as follows. Section 2 reviews lerature and develops hypotheses. Section 3 discusses research design, methods, and data collection. Section 4 presents the descriptive statistics and empirical results. Section 5 discusses and concludes. 2. Lerature and Hypothesis This section reviews lerature and develops hypotheses that are related to earnings informativeness, earnings management, corporate governance functions and changes in regulatory environment. 2.1 Independent Aud Commtee 2.1.1Earnings Informativeness and Independent Aud Commtee There is a long-standing assumption that stronger corporate governance is conducive to better qualy of accounting information (Dey, 2005). However, previous studies do not find consistent evidence on the relation between earnings informativeness and independent aud commtee. Also, previous studies investigate the valuation effect of a majory independent aud commtee rather than a fully independent aud commtee, which is now required to be disclosed under SOX. For example, examining the link between earnings informativeness and corporate governance, Anderson, Deli and Gillan (2003) find earnings informativeness posively related to board independence. Yet, 5

8 they find that aud commtee independence has no incremental power to explain the earnings informativeness of board independence. In contrast, Bryan, Liu and Tiras (2004) find that a fully independent aud commtee is posively related to earnings informativeness over the periods of 1996 through Further, Dey (2005) using relatively small sample finds significant results between earnings informativeness and aud commtee effectiveness in her high agency cost group. In addion, Akhigbe and Martin (2006) show favorable SOX impact on the market valuation in financial service firms when their corporate governance functioning is composed of majory-independent aud commtees and independent boards. After the financial scandals, investors demand accurate and reliable financial information to make informed investment decisions, and subsequently the Act promulgates new rules on the functioning of aud commtees. For example, the Act grants rights to and imposes obligations on a fully independent aud commtee in monoring the financial reporting that the aud commtee has previously not done. Further, a company s aud commtee needs to resolve any disagreement of financial reporting between management and external audor (Klein, 2003). The SEC also encourages firms to have a majory of independent directors board to complement a fully independent aud commtee in ensuring better monoring functions. 5 Moreover, in order to enhance aud commtee independence, aud commtee members are barred from accepting any consulting, advisory or other compensatory fees from the firm. 6 If investors perceive the passage of SOX regulations have strengthened the functions of a fully independent aud commtee, we would expect the earnings 5 Refer to footnote 2. 6 Refer to footnote 2. 6

9 informativeness to be posively related to the disclosure of a fully independent aud commtee after SOX. The alternative hypothesis is stated as: H1a: Earnings informativeness is posively related to the disclosure of a fully independent aud commtee after SOX Earnings management and Independent Aud Commtee Several pre-sox studies (e.g., Klein, 2002; Bryan et al., 2004; Jenkins, 2002) have examined the relation between earnings management and aud commtee effectiveness. Klein (2002) reports a negative relation between discretionary accruals and a majory independent aud commtee. Her results, however, do not support that a fully independent aud commtee has incremental power over the board independence in monoring discretionary accruals. On the other hand, Bryan et al. (2004) find that an effective aud commtee improves the qualy of reported earnings. Likewise, by investigating the joint effects of aud commtee effectiveness and audor independence on earnings management, Jenkins (2002) finds that both effective aud commtee and independent audor migate income-increasing earnings management. After SOX, a fully independent aud commtee seems to play an active role in monoring mangers earnings management behaviors. The legal status of an aud commtee is strengthened after SOX, and the aud commtee is responsible for the appointment of the independent audor. In addion, the Act also directs firms to provide their aud commtee wh independent counsel(s) and other advisors in an effort to assure that the aud commtee is capable of performing s mandatory tasks effectively. If a fully independent aud commtee under SOX is effective in deterring managers 7

10 earnings management behavior, the relation between discretionary accruals and a fully independent aud commtee is expected to be negative. The alternative hypothesis is stated as: H1b: Earnings management is negatively related to the disclosure of a fully independent aud commtee after SOX. 2.2 Financial Experts on Aud Commtee Earnings Informativeness and Financial Experts on Aud Commtee The Act requires that at least one financial expert ss on the firm s aud commtee. The designation of financial experts would prompt investors to demand specific accountabily from board members wh financial expertise. At the same time, the financial experts on the aud commtee would have greater authory to effectively handle issues between management and audors in the process of financial reporting (Mchell, 2003). Previous studies indicate that financial experts on aud commtees affect the effectiveness of firms corporate governance functions. For example, McDaniel, Martin and Maines (2002) investigate the difference between financial experts and financial lerates on the credibily of financial reporting using pre-sox sample. Their results suggest that having financial experts on the aud commtee is likely to change the corporate governance structures and affect the commtee s overall assessment of the financial reports. Bryan et al. (2004) also report that financially lerate aud commtee members increase earnings informativeness. In addion, Akhigbe and Martin (2006) find 8

11 a favorable valuation effect of aud commtee members wh financial expertise in the pre-sox financial service industry. Originally, the proposion for financial experts is to require financial experts having accounting education or experience. The rationale is that aud commtee members would understand the accounting numbers better when they have accounting expertise, which in turn enables them to effectively communicate wh managers and external audors in monoring financial reporting. However, several studies (e.g., Dhaliwal et al., 2006; Carcello et al., 2006; DeFond, Hann, and Hu, 2005) have suggested that the definions of financial experts in Section 407 are too broad to ensure financial reporting qualy. 7 For instance, Defond et al. (2005) show that the market only shows favorably to financial experts wh accounting background or expertise on the aud commtee. We therefore define an aud commtee member as a financial expert if the member has accounting background or experience. If investors perceive that financial experts wh sophisticated accounting knowledge improve the effectiveness of the aud commtee, the earnings informativeness is expected to be posively related to the disclosure of financial experts on aud commtees after SOX. We test the alternative hypothesis stated as follows: H2a: Earnings informativeness is posively related to the disclosure of financial experts wh accounting background or experience on the aud commtee after SOX. 7 Section 407 defines the term financial expert based on (1) the understanding of generally accepted accounting principle and financial statements; (2) experience related to preparation or auding of financial statements; (3) experience related to the application for estimates, accruals, and reserves; (4) experience wh internal control and wh an understanding of aud commtee functions. 9

12 2.2.2 Earnings Management and Financial Experts on Aud Commtee Several studies have examined the relation between financial experts and earnings management. Xie, Davison, and DaDalt (2003) use pre-sox samples to examine whether aud commtee members wh financial background help monor earnings management. They find that the presence of aud commtee members wh corporate or financial background is negatively related to discretionary current accruals. Similarly, Dhaliwal et al. (2006) investigate three types of financial expertise (accounting, finance, and supervisory) on aud commtees and find a posive relation between accruals qualy and accounting experts. However, they do not find significant relation between accruals qualy and non-accounting experts. Using 350 post-sox samples Carcello, Hollingsworth, Klein, and Neal (2006) find that both accounting experts and certain type of non-accounting experts (such as bankers, venture capalists, or CEOs of other firms) reduce earnings management when firms have weaker corporate governance mechanism. Since requires a high degree of accounting sophistication for aud commtee members to perform their task proficiently, financial experts wh high degree of accounting knowledge are expected to improve the aud commtee s effectiveness in monoring the managers earnings management behaviors. We therefore test the effectiveness of financial experts wh accounting background or experience in monoring earnings management after SOX. The alternative hypothesis is stated as: H2b: Earnings management is negatively related to the disclosure of financial experts wh accounting background or experience on the aud commtee after SOX. 10

13 2.3 Dualy of CEO Serving as Chair of the Board Earnings Informativeness and Dualy of CEO Since 1980s, a majory independent board has become the norm in corporate governance. There is an argument whether a CEO should serve as the chair of the board in a majory independent board. On one hand, the dual-role CEO enhances a firm s performance by having a focused direction for the firm s strategies and operations. On the other hand, when corporate insiders other than the CEO are absent from a majory independent board, directors become more dependent upon their link wh the CEO for inside information. As a result, the crical information is often hidden from directors or falsified (Mchell, 2005). Furthermore, the dual role of a CEO (dualy of CEO) implies that the CEO may influence the board on many decisions. Several studies have investigated the relation between earnings qualy and the dualy of CEO. Anderson et al. (2003) find that earnings informativeness is posively related to firms wh separated CEO and chair posions. Also, Dey (2005) finds partial support that the dualy of CEO is negatively related to the credibily of earnings announcements. In addion, Gul and Wah (2002) report that when firms have dominant CEOs, earnings informativeness is posively related to different levels of insider ownership. However, the aforementioned studies eher did not include post-sox samples or did not distinguish the pre-sox from the post-sox samples in their analyses. We argue that after the financial scandals, investors become more suspicious that the dualy of CEO may jeopardize the board s fiduciary duties in monoring a firm s financial reporting. Accordingly, the earnings informativeness will be negatively related 11

14 to the designation of the dual role of CEO after SOX. We test the alternative hypothesis stated as: H3a: Earnings Informativeness is negatively related to the dualy of CEO after SOX Earnings Management and Dualy of CEO The dualy of CEO potentially increases the risk of the CEOs making final decision on financial reporting, which may in turn increase the costs on monoring managerial behaviors of earnings management. Klein (2002) shows that a CEO sting on the compensation commtee increases the possibily of the firm s earnings management. Furthermore, Gul and Wah (2002) report that firms wh dual-role CEOs are more likely to manipulate discretionary accruals when managerial ownership exceeds 25%. After Enron, the Act emphasizes the legal duty of CEOs by requiring them to certify the credibily of firms financial reports. CEOs are also responsible for disclosing any internal control material weaknesses and possible frauds to their aud commtees and external audors. Yet, the issue whether the dualy of CEO is associated wh earnings management after SOX is an empirical question to be addressed. We therefore test the alternative hypothesis is stated as follows: H3b: Earnings management is posively related to the dualy of CEO after SOX Big 4 Audors Prior to SOX, Teoh and Wong (1993) investigate investors perception of the qualy of the Big 8 (then) clients accounting numbers versus that of the non-big 8 clients. They find that Big 8 audors are perceived to provide greater aud qualy, 12

15 thereby ensure greater credibily in accounting information. Jenkins (2002) also finds that independent audors and independent aud commtees complement each other in monoring income-increasing discretionary accruals. The backlash of Arthur Anderson, however, has cast some doubts regarding whether Big 5 (then) accounting firms exercised independently in assuring the credibily of accounting numbers. In addion, there is a concern about Big 5 (then) audors involvement in the provision of both aud and non-aud services to the same companies prior to SOX. As a result of these concerns, the Act brings external audors into firms corporate governance by restricting non-aud services provided by external audors. Section 301 also makes aud commtees directly responsible for resolving disagreement between management and external audors. Since the Act puts more obligations on firms aud commtees, Klein (2003) suggests that Big 4 audors may shift some of their responsibily of monoring financial reporting to the firms aud commtees after the passage of SOX. Yet, whether the effectiveness of the Big 4 audors has changed after SOX is an empirical issue. We argue that if the markets perceive the Big 4 audors to be effective in monoring firms earnings qualy after SOX, a posive (negative) relation between the earnings informativeness (earnings management) and the Big 4 audors in the post-sox period is expected. The alternative hypotheses are as follows: H4a: Earnings informativeness is posively related to the Big 4 audors after SOX. H4b: Earnings management is negatively related to the Big 4 audors after SOX. 13

16 2.5. Corporate Governance Index Corporate governance structures are multi-dimensional. Each function on individual dimensions may substute or complement each other in several aspects to form an optimal governance structure (Dey, 2005). One stream of corporate governance studies (e.g., Larcker, Richardson and Tuna, 2007; Dey, 2005) constructs multidimensional corporate governance metrics to capture the association wh other factors, such as firm performance, earnings informativeness, earnings management, restatements and anti-takeover provisions. For example, Dey (2005) considers a comprehensive set of individual governance variables into 12 factors to measure various corporate governance dimensions. Her results show that the composion and functioning of the board directors, the effectiveness of the aud commtee, and the dualy of CEO are significantly related to reporting credibily for firms wh the highest agency costs. Another stream of corporate governance studies constructs one-dimensional governance metrics by combining a number of variables for a more meaningful analysis in different aspects of governance functions (e.g. Jenkins, 2002; Brown and Caylor, 2006). For example, Jenkins (2002) uses four components to measure aud commtee effectiveness and finds a negative relation between discretionary accruals and aud commtee effectiveness. Since the Act mandates various accounting-related corporate governance functions to restore investors confidence, s effect on corporate governance is multidimensional. Therefore, we propose the use of an aggregated corporate governance score consisting of five accounting-related governance variables as an indicator of firm s overall corporate governance effectiveness. These five variables include a fully independent aud commtee, a majory independent board, financial experts wh 14

17 accounting background or experience on the aud commtee, Big 4 audors, and the separation of CEO and the chair of the board. 8 If investors perceive that these five corporate governance variables together serve as a signal of strong corporate governance, a high aggregated score would be related to high earnings informativeness. Likewise, a high aggregated governance score would be related to low earnings management. The alternative hypotheses, therefore, are stated as follows: H5a: Earnings informativeness is posively related to the corporate governance score after SOX, and H5b: Earnings management is negatively related to the corporate governance score after SOX. 2.6 Pre- versus Post- SOX Corporate Governance Functions Several studies (e.g., Anderson et al., 2003; Romano, 2005; Zhang, 2007) have cast doubts on the effectiveness of new rules on aud commtees and the other corporate governance functions under the SOX and applicable laws. We investigate this issue by comparing the results of our pre-sox and post-sox samples to present evidence on the change of market perception of corporate governance functions after SOX. If there is a magnude increase of relation between earnings informativeness and corporate governance functions in the post-sox period, one could conclude that the SOX improves earnings informativeness condioned on the disclosure of corporate governance information. Likewise, we compare the results of earnings management between the preand post-sox samples. A magnude increase of relation between earnings management 8 Several previous studies (e.g., Jenkins, 2002; Dey, 2005; Xie et al., 2006) use aud commtee meetings to examine the effectiveness of aud commtee in monoring accounting information. See footnote

18 and corporate governance functions in post-sox periods implies that the SOX improves the effectiveness of corporate governance functions in deterring earnings management. By examining the magnude change of the relations, we can also identify any shift of corporate governance functions in the post-sox period. Thus, the alternative hypotheses are stated as: H6a: A magnude increase in relation between earnings informativeness and corporate governance variables in the post-sox period, and H6b: A magnude increase in relation between earnings management and corporate governance variables in the post-sox period. 3. Research Design This section discusses the methods, models, variable definions, and data collection employed to examine the relation between earnings informativeness (earnings management) and a fully independent aud commtee, and the other corporate governance variables. 3.1 Method and Model of Earnings Informativeness and Variable Definions Previous studies (e.g., Gul and Wah, 2002; Anderson et al., 2003; Dey, 2005) have used the modified Teoh and Wong s (1993) methodology to examine the effect of earnings informativeness condioned on the disclosure of corporate governance composion. Earnings informativeness is measured by earnings response coefficient (ERC) which is the correlation between the market cumulative abnormal returns (CAR) and unexpected earnings (UE ). 16

19 This study adopts Anderson et al. s (2003) model for the relation between earnings informativeness and the disclosure of aud commtee independence and other corporate governance variables under SOX. The model is presented as: CAR = α + α UE 0 + β UE 4 + η UE j 1 * AcctExp * β UE 1 ( Control * CompL + β UE ). 5 + β UE 2 * ChrCEO * AudIndep + β UE + β UE 6 3 * Audor * BrdIndep (1) Definions of the variables in model (1) are given in Appendix A. To examine the relation between earnings informativeness and firm s disclosure of s corporate governance functions, we partion the equation (1) as: ( CAR ) = α1 + β1compl ( UE ) + β Audor 6 + η j + β AudIndep ( Control ). + β BrdIndep 3 + β AcctExp 4 + β ChrCEO 5 (2) Thus, the earnings informativeness of firm i is estimated as: 9 ( 1 1 i 2 i 3 i 4 i 5 i 6 i α + β * CompL+ β * AudIndep + β * BrdIndep+ β * AcctExp+ β * ChrCEO+ β * Audor). A posive (negative) β indicates that the particular i th governance variable increases i (decreases) the earnings informativeness by increasing (decreasing) the incremental explanatory power to the relation between abnormal market returns and unexpected earnings at the earnings announcements. Baber and Kang (2002) suggest that the use of the most recent consensus of analysts forecasts reduces the bias between rounded errors and stock spl problems in the calculation of unexpected earnings between actual earnings and analysts forecasts. We follow their suggestions and measure the unexpected earnings as the difference 9 Since our study focuses on the SOX-related corporate governance functions, we did not include the effect of control variables in our discussions. 17

20 between firms actual accounting earnings and the most recent consensus of analysts forecasts. In addion, to exclude stale analyst forecasts, if an analyst s forecast was made 180 days before the earnings announcement date, the analyst s forecast is excluded. Following the guidance given by the SEC and the national exchanges and markets on board 10 and aud commtee independence, we code the firm s aud commtee independence accordingly. 11 In addion, we classify a firm s aud commtee as independent if the firm s non-independent member has resigned from s aud commtee when financial statements are released. 12 Furthermore, most firms in our sample did not disclose their individual director s independence except for their aud commtee members. Firms, however, disclosed whether they followed the applicable listing standards to define a director s independence. We therefore follow the applicable laws and guidelines of the national securies markets to dichotomize a director s independence. We also include control variables in model (1). These control variables include managerial ownership (Mown), market-to-book ratio (MB), market value (MV), and leverage (LEV). Selections of control variables were based on theories and findings of 10 Refer to footnote For example, we classified the aud commtee of DIGI International Inc. as independent based upon the disclosure on s proxy statement: Aud Commtee presently consists of Messrs. Millard (Chairman), Moroz and Williams. All members of the Aud Commtee are independent as that term is defined in the applicable listing standards of The Nasdaq Stock Market and other applicable laws The responsibilies of the Aud Commtee are set forth in the Aud Commtee Charter, a copy of which is included as Exhib A to this Proxy Statement and which was amended by the Company s Board of Directors in November 2002 to comply wh certain requirements of the Sarbanes-Oxley Act of For example, Walt Disney s aud commtee was coded as independent according to the disclosure on s proxy statement. All current members of the Aud Commtee are independent whin the meaning of the listing standards of the New York Stock Exchange. John Bryson served as a member of the Commtee during a portion of the year [fiscal 2002]. 12 Because Mr. Bryson s wife was employed by Lifetime Entertainment Services, in which the Company has an indirect 50% interest, Mr. Bryson may not have been deemed independent under the Exchange s existing rules, as in effect at the time. Accordingly, Mr. Bryson resigned from the Commtee in November

21 previous studies. For instance, Warfield, Wild and Wild (1995) show that managerial ownership is posively related to earnings informativeness but is inversely related to discretionary accruals. Recent studies (e.g., Bushman, Chen, Engel and Smh, 2004; Dey, 2005) have reported that firm size, risk and growth opportunies might affect the magnude of earnings informativeness. In addion, managers are likely to explo earnings management to relieve potential constraints of debt covenants. Therefore, firm s leverage might also affect the effectiveness of corporate governance functions in monoring accounting information. To test the pre-sox relation between earnings informativeness and the disclosure of aud commtee independence and the other corporate governance functions, the variable, Full, replaces CompL in model (1). Full is a dummy variable wh a value of one if the aud commtee was fully independent under the pre-sox definions. Likewise, AudIndep and BrdIndep are classified according to the pre-sox definions. Since there were no requirements for disclosure of financial experts on aud commtees prior to SOX, we classify financial experts based on their education background and experiences as previously defined in the post-sox sample. 3.2 Models of Measuring and Testing Earnings Management Previous studies using the Jones model (Jones, 1991) or the modified Jones model (Dechow et al., 1995) to measure discretionary accruals have been cricized for the problematic heteroskedasticy and the misspecified issues. Recently, Kothari, Leone and Wasley (2005) propose a model that includes an intercept and lag ROA (return on assets) to migate the econometric problems of the Jones and the modified Jones models in 19

22 estimating accruals. We therefore adopt the Kothari et al. s model and use two-dig SIC in the cross-sectional regression to measure the discretionary accruals. First, we use the model to obtain each observed firms post-sox industry estimates in the same two-dig SIC code. The procedures are also applied to the observed firms pre-sox industry estimates. The following estimation model is run by all firms in the same two-dig SIC. TACC = 0 1( δ + δ 1/ AT ) + δ ( Sale ) / AT + δ ( PPE / AT ) δ ROA. (3) Second, we use the observed firms estimated coefficients whout intercept that is obtained from the equation (3) to calculate the nondiscretionary accruals from the modified Jones model wh lagged ROA as follows: NDACC jt = φ (1/ AT 1 + φ ( PPE 3 jt 1 ji ) + φ ( Sale / AT 2 jt 1 ) + φ ROA 4 jt REC. jt 1 jt ) / AT ij 1 (4) Third, the discretionary accruals ( DACC ) of the observed firms are calculated by using jt the total accruals ( TACC ) minus non-discretionary accruals of equation (4) as follows: jt DACC jt = TACC NDACC. (5) jt jt The regression model to examine the post- SOX relation between earnings management, a fully independent aud commtee, and the other corporate governance variables is as follows: DACC = π + µ CompL µ ChrCEO + µ Audindep µ Audor + ξ µ BrdIndep ( Control ). + µ AcctExp 4 (6) Likewise, the pre-sox model is as follows: 20

23 DACC = ϖ + υ FuLL υ ChrCEO + υ Audindep υ Audor + ψ υ BrdIndep ( Control ). + υ AcctExp 4 (7) The corporate governance and control variables in the above models are as previously defined in Appendix A. 3.3 Data Collection We purposely choose the proxy statements for the post-sox samples in this study because the proxy statements include financial reports. Following the Act that was enacted on July 30, 2002, the SEC requires all publicly traded firms to disclose a fully independent aud commtee by January 15 or October 31, 2004, whichever was applicable. Therefore, the first compliant disclosure of a fully independent aud commtee would be found in firms financial reports. Further, we only include firms wh earnings announcement dates after the Act effective date. The post-sox data are retrieved from Compustat, I/B/E/S, and CRSP. First, we retrieved the NYSE/Amex and Nasdaq sample firms wh 2002 financial data available from Compustat. The original sample size was 16,970 firms. We excluded firms which had revenue less than $25 million. 13 The sample size was reduced to 5,221 firms. We further eliminated 764 foreign firms and also eliminated 1,220 utily and financial firms wh two-dig SIC code of 49 and 60. After deleting firms wh missing financial data on Compustat, the sample size was 2, According to the SEC, the definion of small business issuers is firms wh less than $25 million annual revenue. This crerion would exclude most of small business issuers who were required to comply wh the disclosure no later than July 31, Refer to footnote 2 also. 21

24 We then obtained the most recent consensus of analysts earnings forecasts from I/B/E/S. The sample size was subsequently reduced to1, 257 firms wh data available in I/B/E/S. Using the 1,257 firms we retrieved data from CRSP to calculate the three-day cumulative abnormal returns (CAR) around earnings announcement. There were 1,188 firms wh data available from CRSP. The sample size was further reduced to 1,162 firms wh available proxy statements from the SEC s Edgar webse. After eliminating outliers in UE, CAR and financial ratios, the final sample consists of 1,104 firms, which were used to examine the post-sox relation between earnings informativeness and corporate governance functions. To investigate the post-sox relation between earnings management and corporate governance functions, we started wh the 1,104 firms and collected other necessary data from Compustat for the estimation of discretionary accruals. Consequently, we ended up deleting addional 68 firms which were eher in the industry wh less than eight firms, or had estimated discretionary accruals in the top and bottom two percent level. The final sample, thus, consists of 1,036 firms. Similar procedures were followed to select the pre-sox samples for comparison purposes, except that the proxy statements were used to collect the corporate governance information for the pre-sox period. In addion, we only included firms wh earnings announcement date before the Act effective date, July 30, 2002 for the pre-sox period. Thus, the sample firms used to investigate the pre-sox relation between earnings informativeness (earnings management) and corporate governance functions are 1,063 (1,004). The details of data collection procedures are summarized in Appendix B. 22

25 4. Empirical Results 4.1 Descriptive Statistics Table 1 provides the descriptive statistics of the pre- and post- SOX samples. There are 1,063 (1,104) firms in the pre-sox (post-sox) period. The average board size (Brdsz) is 8.30 (8.26) members wh 3.52 (3.49) aud commtee members (Audsz) in the pre-sox (post-sox) sample. Further, the average independent board members (IndBrd) are 5.86 (5.80) that account for 69.73% (69.70%) of total board members (BrdIndep) in the pre-sox (post-sox) period. The average independent aud commtee members (IndAud) are 3.43 in both the pre- and post-sox periods. Moreover, the percentage of independent aud commtee members (AudIndep) is (98.18) whereas the compliant percentage of a fully independent aud commtee wh at least three members (Full/CompL) is (93.21) in the pre-sox (post-sox) sample. The average number of financial experts on aud commtee (Expert) is 0.58 (0.52) which accounts for 16.97% (15.31%) of total aud commtee members (AcctExp) in the pre-sox (post- SOX) period. 14 The percentage of firms having dualy of CEO (ChrCEO) is (60.69), and the percentage of firms having Big4 audors (Audor) is (95.92) in the pre-sox (post-sox) period, respectively. The Wilcoxon signed tests show that among the corporate governance variables Expert, AudIndep, Full/CompL and AcctExp show statistically significant differences between the pre- and post-sox samples 14 The mean (median) number of financial expert is 1.84 (2.00) prior to SOX. The mean (median) number of financial expert is 1.01 (1.00) after SOX according to the definions of Section 407. Consistent wh previous studies (e.g. DeFond et al., 2005; Carcello et al., 2006; Dhaliwal et al, 2006), when we follow the definions of financial expert under Section 407, our regression results show no significant relation between financial expert and earnings informativeness (earnings management). When financial expert is defined as directors having accounting background or experience, however, we find significant relation between financial expert and earnings informativeness (earnings management) in the post-sox period. 23

26 Descriptive statistics on the control variables are also reported in Table 1. The Wilcoxon signed test show that MB, LEV, MV and TA are all significantly different between the pre- and post-sox samples. These differences are mainly contributed by the significant decline in the stock market prices after SOX Table 1 inserted about here Table 2 provides the pre- and post-sox Pearson correlation matrixes. It shows a significantly posive correlation between CAR and UE in both the pre- and post-sox periods. It also shows that the correlation between CAR and UE*CompL is more profound in the post-sox period. Consistent wh our expectation, we find the corporate governance variables are highly correlated wh each other Table 2 inserted about here Pre- versus Post-SOX Earnings Informativeness Table 3 provides the regression results of the pre-sox earnings informativeness on firm-specific corporate governance variables. The results only show a significant relation between earnings informativeness and board independence (BrdIndep, t=1.77 and t=1.65 in models 3 and 4). In contrast, there is no significant relation between earnings informativeness and a fully independent aud commtee (Full), or the percentage of aud commtee independence (AudIndep). The results, however, show a significant relation between earnings informativeness and Big 4 audors (Audor, t=2.22 and t=2.20 in models 1 and 2, respectively) when board independence is excluded from 15 We also ran the descriptive statistics using the 872 firms which were included in both the pre- and post- SOX samples. Similar to the results reported in Table 1, we find significant differences on MB, LEV and MV of these firms in the pre- and post-sox periods. However, TA, which is reported significant at 5% level in Table 1, is found to be insignificant using the 872 firms. To control for the effect of change in these variables on earnings qualy, we therefore include MB, LEV and MV in our regression analyses. 24

27 the regressions. Furthermore, earnings informativeness is not related to the dualy of CEO (ChrCEO). In general, consistent wh previous studies (e.g., Anderson et al., 2003), we find ltle evidence suggesting that pre-sox earnings informativeness is related to corporate governance mechanism other than board independence and Big 4 audors Table 3 inserted about here Table 4 provides the regression results of the post-sox earnings informativeness. The results show a significantly posive relation between earnings informativeness and a fully independent aud commtee (CompL, t=2.19 in model 1). That is, investors perceive the disclosure of a fully independent aud commtee as an indication of better qualy accounting information and react favorably to the firm s earnings announcement. The earnings informativeness related to majory independent aud commtees (AudIndep) is also found to be significantly posive (t=1.95 in model 2). More importantly, earnings informativeness is significantly related to a fully independent aud commtee (CompL) and majory independent board (BrdIndep) in the same regression model (t=1.90 and 1.74 in model 4). These findings support our hypothesis that a fully independent aud commtee provides incremental information effect in addion to the board independence after the passage of SOX Table 4 inserted about here Consistent wh previous studies (e.g., DeFond et al., 2005), this study also finds a significant relation between earnings informativeness and financial experts (AcctExp) on the aud commtee. Furthermore, financial experts and a fully independent aud commtee (CompL) are both significantly and posively related to earnings informativeness (t=2.19 and 1.91 in model 1). Since financial experts wh accounting 25

28 background are able to communicate more effectively wh external audors, our findings suggest that the markets favor firms wh financial experts on a fully independent aud commtee. In addion, Table 4 shows a negative and significant relation between earnings informativeness and the dualy of CEO (ChrCEO) in the post-sox period. Our findings are consistent wh previous studies (e.g., Gul and Wah, 2002; Anderson et al., 2003; Dey, 2005) that CEOs wh dual roles may influence the firms financial reporting, and the markets are unfavorable to firms wh dualy of CEOs. Further, a negative but insignificant relation between earnings informativeness and the Big 4 audors (Audor) is reported in Table 4. This insignificant relation is consistent wh Klein (2003) that after SOX external audors may shift the burden of overseeing the financial reporting to the firm s aud commtee. Also, Table 4 shows a significantly posive relation between earnings informativeness and the aggregate governance score, suggesting that the markets are favorable to firms wh overall strong corporate governance. This finding is consistent wh previous studies that corporate governance mechanisms may substute or complement each other from several dimensions to form an optimal monoring structure (Dey, 2005; Larcker et al., 2007). In summary, comparing the relations between earnings informativeness and the corporate governance variables in the pre- and post-sox periods, we find a magnude change of relations in the variables of Full/CompL, AudIndep, AcctExp, ChrCEO, Audor and Gscore. The magnude change in the relations suggests that there is a change of market perception of aud commtee independence, financial experts on aud 26

29 commtees, separation of CEO and chair of board, Big 4 audors, and a firm s overall corporate governance on the firms credibily of financial reporting after SOX. Our findings provide evidence that the markets perceive the aforementioned corporate governance mechanisms as important indicators of credible financial reporting in the post-sox period. 4.3 Pre- versus Post- SOX Empirical Results of Earnings Management Panel A of Table 5 shows 1,004 pre-sox and 1,036 post-sox firms descriptive statistics of discretionary accruals (DACC, discretionary accruals), total accruals (TACC) and non-discretionary accruals (NDACC). The pre-sox mean of discretionary accruals is 0.019, which is a mean difference between total accruals ( 0.070) and non-discretionary accruals ( 0.051). The post-sox mean of discretionary accruals (DACC, discretionary accruals) is 0.012, which is a mean difference between the total accruals (TACC, 0.056) and non-discretionary accruals (NDACC, 0.044). The Wilcoxon signed tests show that the pre-sox discretionary accruals and total accruals are significantly larger in absolute value than that of the post-sox period (Panel A of Table 5). Our results are consistent wh findings of Cohen, Dey and Lys (2008) which show that earnings management was significantly higher prior to SOX and declined significantly after the passage of SOX. Furthermore, the Pearson correlations in Panel B show no significant correlation between discretionary accruals and a fully independent aud commtee in the pre-sox period, while discretionary accruals are significantly and negatively correlated wh board independence and Big 4 audors in the pre-sox period. By contrast, we find that in the post-sox period, discretionary accruals are significantly and negatively correlated wh a 27

30 fully independent aud commtee, a majory independent aud commtee, an independent board, and financial experts on aud commtees. The association between discretionary accruals and Big 4 audors in the post-sox period, however, is found to be insignificant. There is also no significant correlation between discretionary accruals and the dualy of CEO Table 5 inserted about here Table 6 provides the pre-sox empirical results of earnings management. The regression results show a significantly negative relation between discretionary accruals and board independence (BrdIndep, t= 1.76 and 1.74 in models 3 and 4, respectively). The discretionary accruals, however, are not significantly related to a fully independent aud commtee, majory independent aud commtee or financial experts on aud commtee prior to SOX. In addion, we find no significant relation between the discretionary accruals and the dualy of CEO. The results in Table 6 also show a negative and significant relation between discretionary accruals and Big 4 audors (Audors), indicating that Big 4 audors play a significant role in monoring earnings management prior to SOX. Our findings, nevertheless, do not support the common perception that Big 4 audors who normally provide non-aud services to the same company prior to SOX would influence their independent role in reviewing the firm s financial statements. Furthermore, we find a significantly negative relation between discretionary accruals and an aggregated corporate governance score (Gscore), suggesting that individual corporate governance functions could incrementally complement each other in monoring managerial behaviors of earnings management. 28